Multi-level early warning method and system for intelligent equipment fault and related device

Abstract

The embodiment of the invention provides a multi-level early warning method and system for faults of intelligent equipment and a related device. The method comprises the following steps: receiving edge node data from edge nodes of a laboratory; according to the historical operation data of the plurality of intelligent devices, the environment baseline of the laboratory and the fault records of the plurality of intelligent devices, generating healthy operation thresholds corresponding to the plurality of intelligent devices; performing weighted scoring on the edge node data, and dividing risk levels of the plurality of intelligent devices according to weighted scores and a health operation threshold value; executing an early warning strategy according to the risk level grading; and establishing an association relationship among the fault events of the plurality of intelligent devices, the experiment batch information and the experiment environment change, and dynamically adjusting a healthy operation threshold according to the association relationship. Therefore, through multi-dimensional sensing and dynamic execution of the early warning strategy, active prevention and grading disposal of equipment faults are realized, the real-time performance of safety management of the drug laboratory is effectively improved, and accurate monitoring of early warning of the laboratory is realized.

Description

Technical Field

[0001] This application relates to the field of laboratory safety monitoring technology, and in particular to a multi-level early warning method, system and related device for intelligent equipment failure. Background Technology

[0002] In sophisticated scientific research settings such as pharmaceutical laboratories, equipment safety monitoring and fault early warning systems are crucial for ensuring experimental safety and efficiency. Because experimental equipment and reagents pose risks such as high temperatures, high pressures, and toxic gas leaks, real-time monitoring of multiple parameters is essential.

[0003] Existing single-point alarm systems rely too heavily on manual judgment, triggering alarms only on fixed thresholds. They cannot dynamically adjust response strategies based on the severity of anomalies, leading to operator fatigue or overlooking real risks. Alarm logic is fixed to local controllers, preventing remote updates to algorithm models or expansion with new sensor types. Laboratory renovations require downtime for hardware and driver deployment. Environmental parameters, equipment status, and experimental operation records are stored in a scattered manner, making it difficult to correlate and analyze the root causes of failures, relying on manual experience for troubleshooting. Current technologies are insufficient to meet the demands of real-time, accurate monitoring and dynamic risk-level response in high-throughput, high-risk scenarios in pharmaceutical laboratories. Summary of the Invention

[0004] In view of this, this application provides a multi-level early warning method, system and related device for intelligent equipment failure. By sensing in multiple dimensions and dynamically executing early warning strategies, it realizes proactive prevention and graded handling of equipment failure, effectively improves the real-time performance of drug laboratory safety management, and achieves accurate monitoring of laboratory early warning.

[0005] In a first aspect, embodiments of this application provide a multi-level early warning method for intelligent device faults, applied to the server of a laboratory safety management platform, comprising:

[0006] The system receives edge node data from an edge node in the laboratory. The edge node is connected to multiple sensors installed in the laboratory and on multiple smart devices. The multiple sensors are used to collect monitoring data from the multiple smart devices and experimental environment data from the laboratory. The edge node data is obtained by the edge node after preprocessing the monitoring data and the experimental environment data.

[0007] Based on the historical operating data of the multiple smart devices, the environmental baseline of the laboratory, and the fault records of the multiple smart devices, a health operation threshold corresponding to the multiple smart devices is generated.

[0008] The edge node data is weighted and scored using a random forest algorithm. The risk levels of the multiple smart devices are then classified based on the weighted scores and the health operation threshold. The risk level characterizes the degree of anomaly during the operation of the smart devices. The greater the deviation between the edge node data and the health operation threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk, which reflects a low degree of anomaly; Level 2 risk, which reflects a relatively high degree of anomaly; and Level 3 risk, which reflects a high degree of anomaly.

[0009] The early warning strategy is executed according to the risk level classification to eliminate malfunction events of the multiple smart devices;

[0010] Establish the correlation between the failure events of the multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjust the health operation threshold according to the correlation.

[0011] In one possible embodiment, the step of using a random forest algorithm to weight the edge node data includes: configuring weight parameters for the multi-dimensional data in the edge node data using the random forest algorithm, wherein the weight parameters are used to characterize the degree of influence of the edge node data on the laboratory safety, and the higher the weight parameters, the greater the safety impact of the edge node data on the laboratory; the multi-dimensional data includes at least one of the following: temperature, pressure, vibration spectrum, and operation logs; normalizing the edge node data; and calculating the weighted score based on the weight parameters and the normalized edge node data.

[0012] In one possible embodiment, classifying the risk levels of the multiple smart devices based on the weighted score and the health operation threshold includes: when the weighted score is in the first-level risk score range, determining whether there is a single data point in the edge node data that deviates from the corresponding health operation threshold; if so, determining the smart device corresponding to the weighted score as the first-level risk, wherein the risk levels correspond to risk score ranges respectively; when the weighted score is in the second-level risk score range, determining whether there are two or more related data points in the edge node data that deviate from the corresponding health operation threshold; if so, determining the smart device corresponding to the weighted score as the second-level risk; if not, determining the smart device corresponding to the weighted score as the first-level risk; when the weighted score is in the third-level risk score range, determining whether there is at least one data point in the edge node data whose rate of change exceeds a preset rate; if so, determining the smart device corresponding to the weighted score as the third-level risk; if not, determining the smart device corresponding to the weighted score as the second-level risk.

[0013] In one possible embodiment, the execution of the early warning strategy according to the risk level classification includes: when the risk level is determined to be Level 1 risk, pushing an early warning notification to a mobile device, where the mobile device refers to the communication device of the laboratory personnel, and the early warning notification is used to notify the personnel that a certain device is abnormal and suggest them to go for inspection; when the risk level is determined to be Level 2 risk, activating the audible and visual alarm installed inside the laboratory; and recording the device status when the abnormal device malfunctions and saving it to a fault knowledge base; when the risk level is determined to be Level 3 risk, forcibly cutting off the power supply to the abnormal device and activating the laboratory's emergency system to reduce the degree of harm to the laboratory.

[0014] In one possible embodiment, after executing the early warning strategy according to the risk level classification, the method further includes: acquiring experimental video through an image sensor installed in the laboratory; integrating the alarm timestamp of the early warning strategy, the laboratory video recording, and the operation log of the smart device to generate an interactive timeline, the interactive timeline being used to characterize the correlation between time information and the fault time, for viewing by the maintenance personnel of the laboratory safety management platform; and generating a fault report based on the laboratory video recording and the operation log, outputting the fault report to the laboratory safety management platform, the fault report including the cause of the fault event, the cause being human error and / or equipment defect.

[0015] In one possible embodiment, dynamically adjusting the health operation threshold based on the correlation includes: training a self-correcting model algorithm by combining historical operating data of the multiple smart devices and fault records of the multiple smart devices; and adjusting the health operation threshold through the self-correcting model algorithm so that the health operation threshold is adapted to the working status of the smart devices and the experimental environment of the laboratory.

[0016] In one possible embodiment, the method further includes: encapsulating the control logic of the smart device into a lightweight Docker image file; when a new smart device is added to the laboratory, determining the corresponding Docker image file according to the model of the new smart device and pushing it to the edge node via the MQTT protocol; and downloading the Docker image file through the edge node to establish a communication channel for the new smart device.

[0017] Secondly, embodiments of this application provide a multi-level early warning system for intelligent device faults, including: a receiving unit, a generating unit, a risk assessment unit, an early warning unit, and an adjustment unit; wherein, the receiving unit is used to receive edge node data from an edge node in a laboratory, the edge node being connected to multiple sensors installed in the laboratory and on multiple intelligent devices, the multiple sensors being used to collect monitoring data from the multiple intelligent devices and experimental environment data from the laboratory, the edge node data being obtained by the edge node after preprocessing the monitoring data and the experimental environment data; the generating unit is used to generate health operation thresholds corresponding to the multiple intelligent devices based on the historical operating data of the multiple intelligent devices, the environmental baseline of the laboratory, and the fault records of the multiple intelligent devices; The risk assessment unit is used to perform weighted scoring on the edge node data using a random forest algorithm, and classify the risk levels of the multiple smart devices according to the weighted scores and the healthy operation threshold. The risk level characterizes the degree of anomaly during the operation of the smart devices. The greater the deviation between the edge node data and the healthy operation threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk reflecting low anomaly, Level 2 risk reflecting relatively high anomaly, and Level 3 risk reflecting high anomaly. The early warning unit is used to execute early warning strategies according to the risk levels to eliminate failure events of the multiple smart devices. The adjustment unit is used to establish the correlation between failure events of the multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjust the healthy operation threshold according to the correlation.

[0018] Thirdly, embodiments of this application provide an electronic device, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for performing steps in any method of the first aspect of this application.

[0019] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in any method of the first aspect of this application.

[0020] Fifthly, embodiments of this application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in any method of the first aspect of this application. The computer program product may be a software installation package.

[0021] As can be seen, the multi-level early warning method, system, and related devices for intelligent device failures described above first receive edge node data from edge nodes in the laboratory. These edge nodes are connected to multiple sensors installed within the laboratory and on multiple intelligent devices. These sensors collect monitoring data from the intelligent devices and laboratory environmental data. Second, based on historical operating data from the intelligent devices, the laboratory's environmental baseline, and failure records, health operation thresholds are generated for each intelligent device. Then, a random forest algorithm is used to weight and score the edge node data. Based on the weighted scores and health operation thresholds, the risk levels of the intelligent devices are classified, with each risk level representing the degree of abnormality during operation. Next, early warning strategies are implemented according to the risk level classification to rule out failure events of the intelligent devices. Finally, a correlation is established between failure events of the intelligent devices and experimental batch information and changes in the experimental environment. The health operation thresholds are dynamically adjusted based on this correlation. Thus, through multi-dimensional perception and dynamic execution of early warning strategies, proactive prevention and tiered handling of equipment failures are achieved, effectively improving the real-time nature of drug laboratory safety management and enabling accurate monitoring of laboratory early warnings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the architecture of a laboratory safety management platform provided in an embodiment of this application;

[0024] Figure 2 This is a flowchart illustrating a multi-level early warning method for intelligent device faults provided in an embodiment of this application;

[0025] Figure 3 This is a schematic diagram of a laboratory scene provided in an embodiment of this application;

[0026] Figure 4 This is a schematic diagram illustrating a specific process for classifying risk levels, provided in an embodiment of this application.

[0027] Figure 5 This is a schematic diagram of an interactive timeline provided in an embodiment of this application;

[0028] Figure 6 This is a block diagram of the functional units of a multi-level early warning system for intelligent device faults provided in an embodiment of this application;

[0029] Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0031] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0032] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article indicates that the preceding and following related objects have an "or" relationship.

[0033] In this application's embodiments, "multiple" refers to two or more. In this application's embodiments, "connection" refers to various connection methods, such as direct or indirect connections, to achieve communication between devices; this application's embodiments do not impose any limitations on this.

[0034] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0035] The following describes the relevant content, concepts, meanings, technical issues, technical solutions, and beneficial effects involved in the embodiments of this application.

[0036] LSTM (Long Short-Term Memory) is a special variant of recurrent neural network (RNN) designed to solve the gradient vanishing problem faced by traditional RNNs when processing long sequence data, i.e., the difficulty for the model to capture long-distance dependencies in the sequence.

[0037] Docker is an open-source containerization platform whose core function is to package applications and their dependencies (such as code, runtime, libraries, configuration files, etc.) into lightweight, portable, standardized containers.

[0038] MQTT (Message Queuing Telemetry Transport) is a lightweight instant messaging protocol based on the publish / subscribe model, designed specifically for Internet of Things (IoT) devices.

[0039] JSON (JavaScript Object Notation) is a lightweight text-based data interchange format designed to make data easy for humans to read and write, and easy for computers to parse and generate, when transmitted between machines.

[0040] In sophisticated research settings such as pharmaceutical laboratories, equipment safety monitoring and fault early warning systems are crucial for ensuring experimental safety and efficiency. Due to the risks associated with experimental equipment and reagents, including high temperatures, high pressures, and toxic gas leaks, real-time monitoring of multi-dimensional parameters is essential. Existing single-point alarm systems rely heavily on manual judgment, triggering individual alarms based on fixed thresholds. They cannot dynamically adjust response strategies according to the severity of anomalies, leading to operator fatigue or overlooking real risks. Alarm logic is fixed to local controllers, preventing remote updates to algorithm models or expansion with new sensor types. Laboratory renovations require downtime for hardware and driver deployment. Furthermore, environmental parameters, equipment status, and experimental operation records are stored in a fragmented manner, making it difficult to correlate and analyze the root causes of faults, relying heavily on manual experience for troubleshooting. Current technologies are insufficient to meet the demands of real-time, accurate monitoring and dynamic risk-level response in high-throughput, high-risk pharmaceutical laboratory environments.

[0041] To address the aforementioned issues, this application provides a multi-level early warning method, system, and related device for intelligent equipment failures. By sensing multiple dimensions and dynamically executing early warning strategies, it enables proactive prevention and tiered handling of equipment failures, effectively improving the real-time nature of drug laboratory safety management and achieving accurate monitoring of laboratory early warnings.

[0042] First, the method in this application embodiment is applied to the server of the laboratory safety management platform, combined with Figure 1This application describes a multi-level early warning method for intelligent device faults. Figure 1 This is a schematic diagram of the architecture of a laboratory safety management platform provided in an embodiment of this application, such as... Figure 1 As shown, the laboratory safety management platform 10 includes: a server 110, an edge node 120, and multiple sensors 130. The server 110 is communicatively connected to the edge node 120, and the edge node 120 is connected to the multiple sensors 130. The multiple sensors 130 include, but are not limited to: temperature sensors, pressure sensors, gas concentration sensors, vibration accelerometers, current sensors, etc., and the multiple sensors 130 are connected to the network through the edge node 120.

[0043] First, server 110 receives edge node data from edge node 120 in the laboratory. Edge node 120 is connected to multiple sensors 130 installed in the laboratory and on multiple smart devices. The multiple sensors 130 are used to collect monitoring data from multiple smart devices and experimental environment data from the laboratory. The edge node data is obtained after edge node 120 preprocesses the monitoring data and experimental environment data. Second, server 110 generates health operation thresholds for multiple smart devices based on historical operating data of multiple smart devices, the laboratory's environmental baseline, and fault records of multiple smart devices. Then, server 110 uses a random forest algorithm to analyze the edge nodes. The data is weighted and scored, and the risk levels of multiple smart devices are classified according to the weighted scores and health operation thresholds. The risk level represents the degree of anomaly during the operation of the smart devices. The greater the deviation between the edge node data and the health operation threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk, which reflects a low degree of anomaly; Level 2 risk, which reflects a relatively high degree of anomaly; and Level 3 risk, which reflects a high degree of anomaly. Then, server 110 executes early warning strategies according to the risk level classification to eliminate failure events of multiple smart devices. Finally, server 110 establishes the correlation between failure events of multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjusts the health operation thresholds based on the correlation.

[0044] Specifically, edge node 120 is typically an embedded hardware device, such as an industrial gateway or edge computing box, deployed in a laboratory setting, such as near equipment control cabinets or ventilation systems, directly connecting multiple sensors 130 and server 110. Edge node 120 supports the following protocols: Modbus / RS485: commonly used communication protocols in industrial equipment, such as centrifuges and reactors; ZigBee / Bluetooth: low-power sensor protocols, such as gas sensors and vibration sensors; and OPC UA: a standardized industrial communication protocol supporting cross-vendor device interconnection. This addresses two major problems in traditional solutions where sensors are directly connected to servers: high network latency and data congestion due to large amounts of raw data being directly uploaded to the server, and wasted computing resources as the server must process massive amounts of raw data simultaneously.

[0045] Specifically, multiple sensors 130 are deployed in key parts of the intelligent equipment in the laboratory, as well as in key areas for monitoring the laboratory environment, such as ventilation openings and reagent storage areas. Key parts of the intelligent equipment include: centrifuge bearings, where the bearings are prone to wear due to mechanical stress during high-speed rotation, potentially leading to abnormal vibration, temperature rise, or even shaft breakage or equipment malfunction. Therefore, deploying vibration accelerometers and temperature sensors in the centrifuge bearings can monitor motor current fluctuations and the internal state of the equipment. High-pressure reactor sealing rings are also deployed; these are weak points under high-pressure environments, and aging or improper installation can cause sudden pressure drops / rises and toxic gas leaks. Deploying pressure sensors and gas concentration sensors in the high-pressure reactor sealing rings can monitor the internal temperature gradient and sealing failure characteristics of the pressure vessel.

[0046] The following is combined Figure 2 This application provides a method for multi-level early warning of smart device faults. Figure 2 This is a flowchart illustrating a multi-level early warning method for smart device faults provided in an embodiment of this application, specifically including the following steps:

[0047] Step S210: Receive edge node data from the edge node of the laboratory.

[0048] The edge node is connected to multiple sensors installed in the laboratory and on multiple smart devices. The multiple sensors are used to collect monitoring data from multiple smart devices and experimental environment data from the laboratory. The edge node data is obtained after the edge node performs data preprocessing on the monitoring data and experimental environment data.

[0049] The data preprocessing is performed by the edge nodes and specifically includes: filtering and denoising the raw data to remove interference signals and obtain a first target data packet, such as removing transient interference signals from vibration sensors; configuring multi-dimensional identification tags on the first target data packet and compressing it to obtain a second target data packet, wherein the multi-dimensional identification tags include at least one of the following: device ID, timestamp tag, and location tag; and encapsulating the second target data packet into a standardized JSON format.

[0050] Specifically, please refer to Figure 3 , Figure 3 This is a schematic diagram of a laboratory scene provided in an embodiment of this application, such as... Figure 3 As shown, sensors are installed at key parts of intelligent devices such as centrifuges and high-pressure reactors, and at key areas of the laboratory such as ventilation openings and reagent storage areas, to facilitate the collection of laboratory experimental environment data. For example, a vibration accelerometer (sensor) is installed on the bearing of the centrifuge to monitor the vibration frequency and amplitude and predict bearing loosening or imbalance; a temperature sensor is embedded in the motor or cavity of the centrifuge to monitor temperature changes; a pressure sensor is installed inside the high-pressure reactor to monitor internal pressure; a wind speed sensor is installed at the ventilation opening to monitor wind speed, and a gas concentration sensor is installed at the ventilation opening to monitor volatile gases such as formaldehyde and ethanol.

[0051] Step S220: Based on the historical operating data of multiple smart devices, the environmental baseline of the laboratory, and the fault records of multiple smart devices, generate the health operation thresholds corresponding to multiple smart devices.

[0052] Among them, the environmental baseline refers to the reference range of various parameters of the laboratory environment under normal conditions, which is used to characterize the normal and standardized state of the laboratory.

[0053] Specifically, based on timestamps, equipment operation data, environmental data, and fault records are merged into a unified time series table, as shown in Table 1.

[0054] Table 1

[0055]

[0056] Based on the integrated data, a time-series prediction model, such as LSTM, is trained. The input features are: time-series data from the past 24 hours as the input window (time step = 2880, 1 point every 5 seconds), including: rotational speed, temperature, laboratory temperature, rotational speed-temperature gradient, and the proportion of high-frequency vibration. The output objective is to predict the equipment temperature value for the next hour (continuous prediction, such as outputting the temperature for the next 12 time points, with 5-minute intervals). The specific training process involves: data partitioning: using historical data as the training set (80%) and data from 2024 as the validation set (20%) to ensure model generalization ability. The loss function is the mean squared error (MSE), which measures the deviation between the predicted and actual temperatures. Model optimization uses the Adam optimizer to adjust the LSTM weights and early stopping to prevent overfitting. Finally, the trained LSTM model can output a probability distribution prediction (rather than a single value), for example: under the current conditions of 2200 rpm and 26℃ laboratory temperature, the mean equipment temperature for the next hour is 52℃, with a standard deviation of 1.5℃.

[0057] Specifically, the healthy operation threshold is predicted based on the time-series prediction model, including: 1. Calculation of the baseline value and σ: Baseline value (μ): the mean value predicted by the model (e.g., 52℃), reflecting the "expected temperature" of the equipment under the current operating conditions; Standard deviation (σ): the square root of the variance predicted by the model (e.g., 1.5℃), reflecting the "normal fluctuation range" of the temperature. 2. Definition of healthy operation threshold: Based on the historical data of equipment failure, the healthy operation threshold is set as "baseline value ± 3σ" (covering 99.7% of the normal fluctuation range): Normal range: μ-3σ~μ+3σ (e.g., 52-4.5=47.5℃~52+4.5=56.5℃); Warning range: μ+3σ~μ+5σ (e.g., 56.5℃~60.5℃, Level 1 potential risk); Failure range: >μ+5σ (e.g., >60.5℃, Level 2 moderate risk). 3. Dynamic adjustment mechanism: When the environmental baseline changes (such as the increase in laboratory temperature in summer), the model will be retrained and updated μ and σ. For example, when the laboratory temperature rises from 25℃ to 30℃, the baseline value μ predicted by the model will rise from 52℃ to 55℃, and σ will rise from 1.5℃ to 2℃. The healthy operation threshold will be automatically adjusted to 55-6=49℃~55+6=61℃ to avoid false alarms caused by the increase in environmental temperature.

[0058] Step S230: Use the random forest algorithm to perform weighted scoring on the edge node data, and classify the risk levels of multiple smart devices based on the weighted scores and health operation thresholds.

[0059] The risk level characterizes the degree of anomaly during the operation of intelligent devices. The greater the deviation between edge node data and the healthy operating threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk, reflecting a low degree of anomaly; Level 2 risk, reflecting a relatively high degree of anomaly; and Level 3 risk, reflecting a high degree of anomaly. Specifically, Level 1 risk is a potential risk, characterized by a slight deviation of a single edge node's data from the healthy operating threshold, such as a temperature exceeding the threshold by 5%; Level 2 risk is a moderate risk, characterized by anomalies in data from two or more related edge nodes, such as current fluctuations accompanying pressure increases; and Level 3 risk is an emergency risk, characterized by a rapid deterioration of critical laboratory parameters, such as a 10% increase in gas concentration per second.

[0060] Specifically, in one possible embodiment, a weighted score is calculated using a random forest algorithm for edge node data. This includes: configuring weight parameters for multi-dimensional data in the edge node data using the random forest algorithm. The weight parameters characterize the degree of impact of the edge node data on laboratory safety. The higher the weight parameters, the greater the impact of the edge node data on laboratory safety. The multi-dimensional data includes at least one of the following: temperature, pressure, vibration spectrum, and operation logs; normalizing the edge node data; and calculating a weighted score based on the weight parameters and the normalized edge node data.

[0061] The training of the random forest includes: voting among multiple decision trees: the random forest automatically generates N decision trees, and each tree randomly selects some features (such as only temperature + operation logs) and some samples (such as 70% of the data) for training to avoid overfitting of a single tree; feature importance calculation: when splitting a node, each tree uses Gini impurity or information gain to determine which feature (such as stress) can reduce classification uncertainty to the greatest extent. The more important the feature, the more times it is split in the tree, and finally obtains a higher importance score. Specifically, assuming the input data is: [Temperature = 28℃, Pressure = 1.5MPa, Vibration mean = 150Hz, Operation log = "Modify parameters"], the single decision tree splitting logic is as follows: Level 1: Determine "Pressure > 1.2MPa"; if yes, proceed to the left subtree; Level 2 of the left subtree: Determine "Is the operation log a dangerous instruction"; if yes, classify it as Level 2; Multi-tree integration result: out of 100 trees, 60 are classified as Level 2, 30 as Level 1, and 10 as Level 3, with the final voting result being Level 2.

[0062] In order to ensure that features of different dimensions are transformed to the same scale, the edge node data needs to be normalized. Specifically, the Min-Max normalization method can be used, and the normalization is determined by the following formula:

[0063]

[0064] Where Xmin refers to the minimum value of the feature, Xmax refers to the maximum value of the feature, and Xnorm refers to the normalized value;

[0065] Based on the normalized edge node data, the random forest model training can automatically output the weight parameters of each edge node data, as shown in Table 2. The operation log is represented by coded values, where 0 represents a normal value and 1 represents a dangerous value.

[0066] Table 2

[0067] feature Original value Normalized value ([0,1]) Weight parameters temperature 28℃ 0.4 30％ pressure 1.8MPa 0.7 40％ Vibration spectrum 150Hz 0.9 30％

[0068] For the multi-dimensional parameters collected in real time, a weighted score is calculated according to the weights: Comprehensive score = pressure value × 0.4 + temperature value × 0.3 + vibration mean × 0.2 + operation log code value × 0.1; For example, if the normalized pressure value = 0.7, the normalized temperature value = 0.4, the normalized vibration mean = 0.9, and the operation log code value = 1 (dangerous instruction), then the weighted score = 0.7 × 0.4 + 0.4 × 0.3 + 0.9 × 0.2 + 1 × 0.1 = 0.28 + 0.12 + 0.18 + 0.1 = 0.68.

[0069] As can be seen, in this embodiment, weight parameters are configured for the multi-dimensional data in the edge node data using the random forest algorithm; then, the edge node data is normalized; finally, a weighted score is calculated based on the weight parameters and the normalized edge node data. In this way, the weight parameters are calculated based on the importance of features, avoiding the subjectivity of manually setting weights, and the ensemble learning mechanism of random forest solves the one-sidedness of traditional single-parameter or simple weighting, realizing complex correlation analysis of multi-dimensional parameters such as temperature and pressure, and improving the completeness of risk assessment.

[0070] Specifically, in one possible embodiment, please refer to Figure 4 , Figure 4 This is a schematic diagram illustrating a specific process for classifying risk levels, as provided in an embodiment of this application. Figure 4 As shown, the risk levels of multiple smart devices are classified based on weighted scores and health operation thresholds, including the following steps:

[0071] Step S410: Determine the risk score range corresponding to the weighted score.

[0072] If the weighted score is within the first-level risk score range, then proceed to step S420: determine whether there is a single data point in the edge node data that deviates from the corresponding healthy operation threshold; if not, then determine that the smart device corresponding to the current weighted score is not at risk; if so, then proceed to step S430.

[0073] S430, the smart device corresponding to the weighted score is identified as a Level 1 risk.

[0074] The risk levels correspond to risk score ranges.

[0075] When the weighted score is within the second-level risk score range, proceed to step S440: determine whether there are two or more related data in the edge node data that deviate from the corresponding healthy operation threshold; if not, proceed to step S430; if so, proceed to step S450.

[0076] S450, the smart device corresponding to the weighted score is identified as a level 2 risk.

[0077] When the weighted score is in the third-level risk score range, proceed to step S460: determine whether there is at least one data change rate in the edge node data that exceeds the preset rate; if there is, proceed to step S470; if there is no such data, proceed to step S450.

[0078] S470, the smart device corresponding to the weighted score is determined to be at level three risk.

[0079] Specifically, Level 1 risk is a potential risk, characterized by a slight deviation of a single edge node's data from the healthy operating threshold, such as a temperature exceeding the threshold by 5%; Level 2 risk is a moderate risk, characterized by abnormal data from two or more related edge nodes, such as current fluctuations accompanying pressure increases; and Level 3 risk is an emergency risk, characterized by a rapid deterioration of critical laboratory parameters, such as a 10% increase in gas concentration per second.

[0080] Step S240: Implement the early warning strategy according to the risk level classification to eliminate malfunction events of multiple smart devices.

[0081] In one possible embodiment, the early warning strategy is implemented according to risk level classification, including: when the risk level is determined to be Level 1, pushing an early warning notification to a mobile device, which refers to the communication device of the laboratory personnel, and the early warning notification is used to inform the personnel that a certain device is abnormal and suggest that they go to check it; when the risk level is determined to be Level 2, activating the audible and visual alarm installed inside the laboratory; and recording the device status when the abnormal device malfunctions and saving it to the fault knowledge base; when the risk level is determined to be Level 3, forcibly cutting off the power supply of the abnormal device and activating the laboratory's emergency system to reduce the degree of harm to the laboratory.

[0082] Specifically, the server, through random forest algorithm analysis, found that the centrifuge bearing temperature exceeded the baseline value by 5% for 10 consecutive minutes (the baseline value is 25℃, and the current temperature is 26.3℃); the comprehensive weighted score was 0.45 (below the level 2 risk score range of 0.5), and it was judged as a level 1 risk. The server sent an alert notification to the experimenter's mobile device application (APP) via the MQTT protocol. The notification content could include: device ID, risk level information, specific abnormal data, and time information. After receiving the notification, the experimenter could view the device's real-time data curves and historical trends (such as a temperature fluctuation graph over the past hour) through the mobile device application.

[0083] Specifically, after 2 hours, the system monitored the following: the temperature continued to rise to 28℃ (12% above the baseline); an abnormal peak appeared in the vibration spectrum (200Hz, exceeding the normal range of 150Hz±20Hz); the operation log showed that the experimenter manually adjusted the rotation speed parameters at 14:25; the comprehensive score was 0.68 (falling into the level 2 risk score range of 0.5~0.8), and it was judged as a level 2 risk. Audible and visual alarm activation: The server sent a command to the edge node, which triggered the audible and visual alarm in the laboratory via a relay, emitting an 85dB buzzer + red flash for 30 seconds; simultaneously, the alarm display showed: abnormal temperature + vibration. Equipment status snapshot recording was also performed: While triggering the alarm, the edge node collected and stored the following data in the fault knowledge base: real-time parameters: temperature 28℃, vibration acceleration 2.3g, current 1.8A; operation log: "14:25 rotation speed adjusted from 2000rpm to 2500rpm"; environmental data: current laboratory humidity 65%, ventilation volume 1000m³ / h. 3 / h; where the data is stored in the form of a JSON file with a timestamp, named "L-001_202505241432_level2.json", where L-001 is the device ID information.

[0084] Specifically, if the aforementioned Level 2 risk is not addressed in a timely manner, the system further monitors the following: a sudden temperature rise to 35°C (40% above the baseline value) with a rate of increase of 2°C per minute; a surge in current to 2.5A (20% above the rated current); and a rapid increase in acetone concentration detected by the gas sensor (5 ppm per second), indicating a potential seal leak. The overall score is 0.92 (≥ Level 3 risk score range 0.8), classifying it as a Level 3 risk. Forced power cut-off: The server sends a command to the centrifuge power controller via the edge node, triggering a hardware relay to cut off the main power supply. The cut-off action is completed in <500ms. Simultaneously, a command is sent to the laboratory power monitoring system to record the power outage time and equipment status. Emergency ventilation system activation: The server calls the ventilation system API to increase the ventilation volume to the maximum level (3000 m³ / h) and activate the activated carbon filter. The ventilation system feeds back the execution status to the server, forming a closed-loop verification.

[0085] As can be seen, in this embodiment, by implementing a multi-level early warning strategy according to risk level classification, a phased response is achieved, an early warning scheme that matches the actual urgency of risk is implemented, and a multi-level and multi-dimensional laboratory protection barrier is constructed, which effectively improves the real-time performance of drug laboratory safety management and the comprehensiveness and reliability of safety protection.

[0086] Step S250: Establish the correlation between fault events of multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjust the health operation threshold according to the correlation.

[0087] Specifically, abnormal equipment events, such as motor overheating, are correlated with experimental batch information, reagent formulations, operators, and environmental fluctuations, such as sudden increases in humidity, to uncover potential risk patterns, such as "an operator did not turn on the dehumidifier during an experiment, resulting in a 40% increase in the failure rate."

[0088] As can be seen, in this embodiment, firstly, edge node data from the laboratory's edge nodes is received. These edge nodes are connected to multiple sensors installed within the laboratory and on multiple intelligent devices. These sensors collect monitoring data from the intelligent devices and laboratory environmental data. Secondly, based on historical operational data from the intelligent devices, the laboratory's environmental baseline, and fault records, health operation thresholds are generated for each intelligent device. Then, a random forest algorithm is used to weight and score the edge node data. Based on the weighted scores and health operation thresholds, the risk levels of the intelligent devices are categorized, with each risk level representing the degree of anomaly during device operation. Next, early warning strategies are implemented according to the risk level classification to eliminate fault events from the intelligent devices. Finally, a correlation is established between fault events from the intelligent devices and experimental batch information and changes in the experimental environment. The health operation thresholds are dynamically adjusted based on this correlation. Thus, through multi-dimensional perception and dynamic execution of early warning strategies, proactive prevention and tiered handling of equipment faults are achieved, effectively improving the real-time nature of drug laboratory safety management and enabling accurate monitoring of laboratory early warnings.

[0089] In one possible embodiment, after implementing the early warning strategy according to the risk level classification, the method further includes: acquiring experimental videos through image sensors installed in the laboratory; integrating the alarm timestamps of the early warning strategy, laboratory video recordings, and operation logs of smart devices to generate an interactive timeline, which is used to characterize the correlation between time information and fault time for viewing by the maintenance personnel of the laboratory safety management platform; and generating a fault report based on the laboratory video recordings and operation logs, outputting the fault report to the laboratory safety management platform, the fault report including the cause of the fault event, which is human error and / or equipment defect.

[0090] The image sensor can be a camera or other device, installed on the ceiling of the laboratory, with a field of view covering the equipment operation area (such as the button area of ​​the reactor), key equipment (such as the centrifuge display screen), and environmentally sensitive points (such as the ventilation opening), ensuring comprehensive monitoring of equipment operation, environmental changes, and personnel behavior.

[0091] The server obtains time-related information from three data sources, including alarm timestamps from the risk classification module (e.g., triggering a level-two risk alarm at 14:32:15 on 2020-05-26), experimental video timestamps from the image sensor (e.g., video clip start time 14:30:00 on 2020-05-26, end time 14:35:00), and operation log timestamps from the smart device (e.g., operator A executing the "close pressure relief valve" command at 14:31:40 on 2020-05-26). The server uses the alarm time (14:32:15) as the baseline point, and extracts a time window of 3 minutes before (14:29:15) and 2 minutes after (14:34:15). It then aligns the video clips (14:29:15-14:34:15), operation logs (time points such as 14:31:40 and 14:32:00), and the alarm event (14:32:15) to generate an interactive timeline.

[0092] Specifically, please refer to Figure 5 , Figure 5 This is a schematic diagram of an interactive timeline provided in an embodiment of this application. The interactive timeline is displayed on the display device of the laboratory safety management platform as follows: Figure 5 As shown, the interactive timeline interface 50 includes elements such as a timeline 501, a video preview window 502, an operation log label 503, and an alarm event marker 504. For example, the video preview window 502 allows staff on the laboratory safety management platform to click on any time point on the timeline, such as 14:31:40, which automatically jumps to the corresponding frame in the video, showing the scene of operator A pressing the button to close the pressure relief valve. The operation log label 503 on the interactive timeline interface 50 displays the command operation type, such as "dangerous command" or "normal start command," when the staff hovers the mouse over it. It can also display detailed information about the operation command, such as modifying parameter xx to 0.9. The alarm event marker 504 is used to mark "Level 2 Risk Alarm: Pressure Increase + Current Fluctuation" at the 14:35 time position on the timeline 501. Staff can click or hover the mouse to view the alarm details, such as pressure value 1.5MPa → 1.8MPa and current value 0.6A → 0.8A.

[0093] The server determines the cause of the fault using the following logic: 1. Human error: If there is video recording of non-compliant operation (e.g., operator A operating the valve without protective gloves) or unauthorized instructions (e.g., lack of permission to modify the pressure threshold) within 5 minutes prior to the alarm in the timeline, it is marked as human error; 2. Equipment defect: If the timeline shows normal operation logs (e.g., instructions conform to procedures), but equipment parameters (e.g., pressure) still fluctuate abnormally (vibration spectrum shows abnormal bearing noise), it is marked as equipment defect based on historical maintenance records (e.g., the centrifuge had its aged bearing replaced 3 months ago). Based on the above fault cause analysis, a fault report is output: a PDF fault report is generated, including basic information (laboratory ID, alarm time, equipment model); a timeline screenshot (with key time points marked); a conclusion regarding the fault cause, such as human error or equipment defect; and improvement suggestions, such as strengthening operator training or arranging equipment maintenance.

[0094] As can be seen, in this embodiment, by integrating alarm timestamps of the early warning strategy, laboratory video recordings, and operation logs of smart devices through video acquisition, an interactive timeline is generated. Furthermore, a fault report is generated based on the video information and operation logs. This solves the shortcomings of traditional laboratory equipment faults, such as data being independent and fragmented and responsibility being difficult to define. It provides a visual interactive interface for laboratory safety management, enabling more accurate and faster location of equipment fault causes and effectively reducing equipment damage rates.

[0095] In one possible embodiment, dynamically adjusting the health operation threshold based on the correlation includes: training a self-correcting model algorithm by combining historical operating data and fault records of multiple smart devices; and adjusting the health operation threshold through the self-correcting model algorithm so that the health operation threshold is adapted to the working status of the smart devices and the experimental environment of the laboratory.

[0096] Specifically, historical operating data of 10 intelligent devices in the laboratory, including centrifuges, high-pressure reactors, and ventilation systems, were collected. This included data on temperature, pressure, and vibration frequency over the past 12 months, with a daily sampling interval of 1 minute. Fault records for each device were also collected, such as cases like "centrifuge shutdown due to excessive vibration caused by bearing wear" and "reactor pressure surge due to seal failure." Each record was associated with environmental parameters at the time of the fault, such as laboratory temperature, humidity, and ventilation volume. The equipment operating data was normalized (e.g., temperature range [20℃, 40℃] → [0, 1]). Key characteristics were marked on the fault records (e.g., "bearing wear" corresponds to a vibration spectrum peak > 150Hz, "seal failure" corresponds to a pressure jump slope > 0.5MPa / s).

[0097] Specifically, the training process of the self-correcting model algorithm includes: using a gradient boosting tree as the self-correcting model; input features include: equipment type (centrifuge / reactor, etc., uniquely thermally encoded); historical operating parameters (mean temperature, pressure variance, vibration frequency); environmental parameters (season, laboratory ventilation level); and fault type labels (bearing wear / seal failure, etc., a total of 5 categories). Dynamic adjustment includes: the server sending instructions to edge nodes via the MQTT protocol to request the latest 7-day operating data and current environmental parameters of each device (e.g., laboratory temperature is 28℃ in summer mode); the self-correcting model outputs adjustment suggestions (e.g., "the vibration threshold of the centrifuge is increased from 80dB to 88dB" and "the lower limit of the pressure threshold of the reactor is decreased from 0.8MPa to 0.7MPa"); the edge nodes receive the threshold update packets, synchronously update the local alarm rule base via the Modbus protocol, and return a confirmation receipt to the server.

[0098] As can be seen, in this embodiment, by dynamically adapting to equipment aging and environmental changes, the healthy operation threshold is dynamically adjusted, so that the healthy operation threshold changes from a static fixed value to a dynamic parameter that drifts with the equipment life cycle and the environment, thereby improving the intelligence level of laboratory safety monitoring and enhancing the reliability of safety early warning.

[0099] In one possible embodiment, the method further includes: encapsulating the control logic of the smart device into a lightweight Docker image file; when a new smart device is added to the laboratory, determining the corresponding Docker image file according to the model of the new smart device and pushing it to the edge node via the MQTT protocol; and downloading the Docker image file through the edge node to establish a communication channel for the new smart device.

[0100] Each image file in the driver package contains: necessary dependency libraries for operation, such as a Python environment and communication protocol plugins; and API interfaces, such as the / emergency_stop interface for calling the emergency stop function, for external systems to call. The image size is kept under 50MB to ensure fast loading on edge nodes.

[0101] In traditional solutions, when adding new smart devices to the laboratory, such as a new carbon dioxide sensor, manual on-site installation of drivers and configuration of communication parameters is required, taking several hours and sometimes even causing downtime. In this embodiment, the administrator enters the new device model, such as "carbon dioxide sensor A-100," into the system backend. The cloud automatically matches the corresponding driver image from the image repository based on the device fingerprint, and then airdrops the image to the laboratory's edge gateway via the MQTT protocol. After receiving the image, the edge node establishes a communication channel with the new sensor via the OPC UA protocol.

[0102] As can be seen, in this embodiment, the control logic of the smart device is encapsulated into a lightweight Docker image file. Based on the new smart device model, the corresponding Docker image file is determined and pushed to the edge node via the MQTT protocol. The edge node then downloads the Docker image file, establishing a communication channel for the new smart device. This API-izes the complex internal logic of the device, enabling flexible and seamless integration between the smart device and the laboratory system. This eliminates the need for laboratory downtime for deployment, improves the scalability of laboratory equipment, and effectively enhances the convenience and efficiency of deploying new functional devices in the laboratory.

[0103] In other possible embodiments, this solution may also include displaying a multi-level monitoring interface on the laboratory safety management platform, outputting information of each device through the server, displaying the monitoring status of the devices in real time on the monitoring interface, and using a three-color warning system (green / yellow / red) to dynamically map the device operating status. For example, green: all monitoring parameters are in the healthy range (e.g., temperature ≤ 25℃, pressure ≤ 1.0MPa); yellow: a single parameter is close to the threshold (e.g., temperature reaches 28℃, threshold 30℃), the device icon flashes yellow, indicating "potential risk: temperature is 10% too high"; red: triggering a level 2 risk or higher alarm (e.g., pressure suddenly rises above 1.5MPa), the device icon turns red and the border is highlighted, displaying the warning "emergency risk: please handle immediately".

[0104] It can also display a root cause analysis dashboard: based on the past 30 days of failure data, it automatically generates interactive charts (bar chart + pie chart): the bar chart shows the ranking of causes (e.g., sensor drift 32%, operational error 28%, equipment aging 18%, environmental fluctuation 12%, other 10%); the pie chart drills down by dimension (e.g., "operational error" is broken down into "failure to close valves according to procedures 45%, accidental parameter touch 30%, failure to wear protective equipment 25%)); it supports time filtering (week / month / quarter) to compare the changing trends of causes in different periods (e.g., the proportion of "environmental fluctuation" increases to 20% in summer).

[0105] In addition, a remote control center is set up in the laboratory safety management platform to realize fully online operation. Specifically, this includes: a drive management module that lists edge nodes and deployed images (such as "Edge Gateway A: Robotic Arm Emergency Stop Drive V1.2, Vibration Calibration Drive V2.0"); it also supports batch selection of nodes, selecting new images via drop-down menus (such as "Carbon Dioxide Sensor Drive V3.0"), and clicking the "Push" button to trigger remote deployment, with a progress bar displaying the transmission status in real time (45MB / 50MB transferred); after deployment, a log is automatically generated (time: 15:05:23, status: success, image hash value: a1b2c3), and supports exporting to CSV or online keyword search (such as "failure" or "timeout"). Alarm rule adjustment module: Visual threshold editor. Taking temperature as an example, the alarm threshold can be adjusted via a slider (current threshold 30℃ → temporarily adjusted to 32℃). The affected area after adjustment is displayed in real time on the right. Area blocking function: temporarily disable alarms in a certain area (valid for 4 hours) to avoid frequent alarm interference during experimental debugging. Operation auditing: all remote adjustment operations are automatically recorded to the backend (operator: xx, time: 15:10, adjustment content: temperature threshold +2℃). It supports hierarchical permission, such as administrators can make global adjustments, while ordinary users can only view them.

[0106] As can be seen, in this embodiment, real data is fed back to the operation and maintenance personnel of the laboratory safety management platform in real time through a multi-level monitoring interface, transforming the passive response of the traditional operation and maintenance mode into proactive dynamic optimization, effectively improving operation and maintenance efficiency, and enhancing system flexibility and security.

[0107] The above primarily describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, mobile electronic devices include corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0108] This application embodiment can divide the electronic device into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0109] and Figure 2 The implementation is consistent with the previous one; please refer to [link / reference]. Figure 6 , Figure 6 This is a functional unit block diagram of a multi-level early warning system for intelligent device faults provided in this application embodiment. The multi-level early warning system 60 includes: a receiving unit 610, a generating unit 620, a risk assessment unit 630, an early warning unit 640, and an adjustment unit 650. The receiving unit 610 receives edge node data from an edge node in the laboratory. The edge node is connected to multiple sensors installed in the laboratory and on multiple intelligent devices. The multiple sensors collect monitoring data from multiple intelligent devices and experimental environment data from the laboratory. The edge node data is obtained after the edge node preprocesses the monitoring data and experimental environment data. The generating unit 620 generates multiple... The system includes a health operation threshold for each smart device; a risk assessment unit 630, which uses a random forest algorithm to weight and score edge node data, and classifies the risk levels of multiple smart devices based on the weighted scores and health operation thresholds. The risk level characterizes the degree of anomaly during the operation of the smart devices. The greater the deviation between the edge node data and the health operation threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk reflecting a low degree of anomaly, Level 2 risk reflecting a relatively high degree of anomaly, and Level 3 risk reflecting a high degree of anomaly; an early warning unit 640, which executes early warning strategies according to the risk level classification to eliminate failure events of multiple smart devices; and an adjustment unit 650, which establishes the correlation between failure events of multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjusts the health operation thresholds based on the correlation.

[0110] In one possible embodiment, regarding the use of a random forest algorithm to weight the edge node data, the risk assessment unit 630 is specifically configured to: configure weight parameters for the multi-dimensional data in the edge node data using the random forest algorithm, wherein the weight parameters characterize the degree of impact of the edge node data on laboratory safety, and the higher the weight parameters, the greater the impact of the edge node data on laboratory safety; the multi-dimensional data includes at least one of the following: temperature, pressure, vibration spectrum, and operation logs; normalize the edge node data; and calculate a weighted score based on the weight parameters and the normalized edge node data.

[0111] In one possible embodiment, regarding the classification of risk levels for multiple smart devices based on weighted scores and health operation thresholds, the risk assessment unit 630 is specifically configured to: when the weighted score is within the first-level risk score range, determine whether there is a single data point in the edge node data that deviates from the corresponding health operation threshold; if so, determine that the smart device corresponding to the weighted score is at first-level risk, wherein the risk levels correspond to risk score ranges respectively; when the weighted score is within the second-level risk score range, determine whether there are two or more related data points in the edge node data that deviate from the corresponding health operation thresholds; if so, determine that the smart device corresponding to the weighted score is at second-level risk; if not, determine that the smart device corresponding to the weighted score is at first-level risk; when the weighted score is within the third-level risk score range, determine whether there is at least one data point in the edge node data whose rate of change exceeds a preset rate; if so, determine that the smart device corresponding to the weighted score is at third-level risk; if not, determine that the smart device corresponding to the weighted score is at second-level risk.

[0112] In one possible embodiment, regarding the implementation of the early warning strategy according to risk level classification, the early warning unit 640 is specifically configured to: when the risk level is determined to be Level 1, push an early warning notification to a mobile device, where the mobile device refers to the communication device of the laboratory personnel, and the early warning notification is used to notify the personnel that a certain device is abnormal and suggest them to go for inspection; when the risk level is determined to be Level 2, activate the audible and visual alarm installed inside the laboratory; and record the device status when the abnormal device malfunctions and save it to the fault knowledge base; when the risk level is determined to be Level 3, forcibly cut off the power supply of the abnormal device and activate the laboratory's emergency system to reduce the degree of harm to the laboratory.

[0113] In one possible embodiment, after executing the early warning strategy according to the risk level classification, the early warning unit 640 is further configured to: acquire experimental videos through image sensors installed in the laboratory; integrate the alarm timestamps of the early warning strategy, laboratory video recordings, and operation logs of smart devices to generate an interactive timeline, which is used to characterize the correlation between time information and fault time for viewing by the maintenance personnel of the laboratory safety management platform; and generate a fault report based on the laboratory video recordings and operation logs, output the fault report to the laboratory safety management platform, and include the cause of the fault event, which is human error and / or equipment defects.

[0114] In one possible embodiment, regarding the dynamic adjustment of the health operation threshold based on the correlation, the adjustment unit 650 is specifically used to: train a self-correcting model algorithm by combining historical operating data of multiple smart devices and fault records of multiple smart devices; and adjust the health operation threshold through the self-correcting model algorithm so that the health operation threshold is adapted to the working status of the smart devices and the experimental environment of the laboratory.

[0115] In one possible embodiment, the multi-level early warning system 60 is further configured to: encapsulate the control logic of the smart device into a lightweight Docker image file; when a new smart device is added to the laboratory, determine the corresponding Docker image file according to the model of the new smart device and push it to the edge node via the MQTT protocol; and download the Docker image file through the edge node to establish a communication channel for the new smart device.

[0116] It is understood that since the method embodiments and the device embodiments are different presentations of the same technical concept, the content of the method embodiment section in this application should be adapted to the device embodiment section in a synchronous manner, and will not be repeated here.

[0117] Figure 7 This is a structural block diagram of an electronic device provided in an embodiment of this application. For example... Figure 7 As shown, electronic device 700 may include one or more components: a processor 701 and a memory 702 coupled to the processor 701, wherein the memory 702 may store one or more computer programs, which may be configured to implement the methods described in the examples above when executed by one or more processors 701. Electronic device 700 may be as follows: Figure 1 The server 110 in the laboratory safety management platform 10 shown.

[0118] Processor 701 may include one or more processing cores. Processor 701 connects to various parts within the electronic device 700 using various interfaces and lines, and performs various functions and processes data of the electronic device 700 by running or executing instructions, programs, code sets, or instruction sets stored in memory 702, and by calling data stored in memory 702. Optionally, processor 701 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). Processor 701 may integrate one or more of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. It is understood that the aforementioned modem may also not be integrated into processor 701, but may be implemented separately through a communication chip.

[0119] The memory 702 may include random access memory (RAM) or read-only memory (ROM). The memory 702 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 702 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), and instructions for implementing the various method examples described above. The data storage area may also store data created during the use of the electronic device 700.

[0120] It is understood that the electronic device 700 may include more or fewer structural elements than those shown in the above block diagram, such as a power module, physical buttons, WiFi (Wireless Fidelity) module, speaker, Bluetooth module, sensor, etc., without limitation.

[0121] This application also provides a computer storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements some or all of the steps of any of the methods described in the above method embodiments.

[0122] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments.

[0123] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0124] In the several embodiments provided in this application, it should be understood that the disclosed methods, apparatuses, and systems can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for example, the division of units is merely a logical functional division, and there may be other division methods in actual implementation; for example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0125] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can be physically comprised separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware or in the form of hardware plus software functional units.

[0127] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, volatile memory, or non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM), etc., which are various media capable of storing program code.

[0128] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can easily conceive of variations or substitutions without departing from the spirit and scope of the present invention, and various modifications and alterations can be made, including combinations of the different functions and implementation steps described above, as well as software and hardware implementation methods, all of which are within the protection scope of the present invention.

Claims

1. A multi-level early warning method for intelligent device faults, characterized in that, The servers used in the laboratory safety management platform include: The system receives edge node data from an edge node in the laboratory. The edge node is connected to multiple sensors installed in the laboratory and on multiple smart devices. The multiple sensors are used to collect monitoring data from the multiple smart devices and experimental environment data from the laboratory. The edge node data is obtained by the edge node after preprocessing the monitoring data and the experimental environment data. Based on the historical operating data of the multiple smart devices, the environmental baseline of the laboratory, and the fault records of the multiple smart devices, a health operation threshold corresponding to the multiple smart devices is generated. The edge node data is weighted and scored using a random forest algorithm. The risk levels of the multiple smart devices are then classified based on the weighted scores and the health operation threshold. The risk level characterizes the degree of anomaly during the operation of the smart devices. The greater the deviation between the edge node data and the health operation threshold, the higher the degree of anomaly. The risk levels include: Level 1 risk, which reflects a low degree of anomaly; Level 2 risk, which reflects a relatively high degree of anomaly; and Level 3 risk, which reflects a high degree of anomaly. The early warning strategy is executed according to the risk level classification to eliminate malfunction events of the multiple smart devices; Establish the correlation between the failure events of the multiple smart devices and experimental batch information and changes in the experimental environment, and dynamically adjust the health operation threshold according to the correlation.

2. The method according to claim 1, characterized in that, The step of using a random forest algorithm to weight and score the edge node data includes: A random forest algorithm is used to configure weight parameters for the multi-dimensional data in the edge node data. The weight parameters are used to characterize the degree of influence of the edge node data on the laboratory safety. The higher the weight parameters are, the greater the safety impact of the edge node data on the laboratory. The multi-dimensional data includes at least one of the following: temperature, pressure, vibration spectrum and operation log. The edge node data is normalized. The weighted score is calculated based on the weight parameters and the normalized edge node data.

3. The method according to claim 1, characterized in that, The step of classifying the risk levels of the multiple smart devices based on the weighted score and the health operation threshold includes: When the weighted score is within the first-level risk score range, it is determined whether there is a single data point in the edge node data that deviates from the corresponding healthy operation threshold. If it exists, the smart device corresponding to the weighted score is determined to be the first-level risk, wherein the risk level corresponds to the risk score range; When the weighted score is within the secondary risk score range, it is determined whether there are two or more related data in the edge node data that deviate from the corresponding healthy operation threshold. If it exists, the smart device corresponding to the weighted score is determined to be the secondary risk; if it does not exist, the smart device corresponding to the weighted score is determined to be the primary risk. When the weighted score is in the third-level risk score range, it is determined whether there is at least one data change rate in the edge node data that exceeds the preset rate. If it exists, the smart device corresponding to the weighted score is determined to be the third-level risk; if it does not exist, the smart device corresponding to the weighted score is determined to be the second-level risk.

4. The method according to claim 3, characterized in that, The implementation of the early warning strategy according to the risk level classification includes: When the risk level is determined to be Level 1 risk, a warning notification is pushed to the mobile device, which refers to the communication device of the laboratory personnel. The warning notification is used to notify the laboratory personnel that a certain device is abnormal and to suggest that they go to check it. When the risk level is determined to be Level 2, the audible and visual alarm installed inside the laboratory is activated; and the equipment status when the abnormal equipment malfunctions is recorded and saved to the fault knowledge base. When the risk level is determined to be Level 3, the power supply to the abnormal equipment is forcibly cut off, and the emergency system of the laboratory is activated to reduce the degree of harm to the laboratory.

5. The method according to claim 4, characterized in that, After implementing the early warning strategy according to the risk level classification, the method further includes: Experimental videos were acquired using image sensors installed in the laboratory. By integrating the alarm timestamps of the aforementioned early warning strategy, the laboratory video recordings, and the operation logs of the smart devices, an interactive timeline is generated. This interactive timeline represents the correlation between time information and the fault time, for viewing by the maintenance personnel of the laboratory safety management platform; and... Based on the laboratory video recording and the operation log, a fault report is generated and output to the laboratory safety management platform. The fault report includes the cause of the fault event, which is human error and / or equipment defect.

6. The method according to any one of claims 1-5, characterized in that, The step of dynamically adjusting the health operation threshold based on the correlation includes: By combining the historical operating data and fault records of the multiple smart devices, a self-correcting model algorithm is trained. The self-correcting model algorithm is used to adjust the health operation threshold so that the health operation threshold is adapted to the working status of the smart device and the experimental environment of the laboratory.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: The control logic of the smart device is encapsulated into a lightweight Docker image file; When a new smart device is added to the laboratory, the corresponding Docker image file is determined according to the model of the new smart device and pushed to the edge node via the MQTT protocol; The Docker image file is downloaded through the edge node to establish a communication channel for the new smart device.

8. A multi-level early warning system for intelligent device faults, characterized in that, include: The system comprises a receiving unit, a generating unit, a risk assessment unit, an early warning unit, and an adjustment unit; among which, The receiving unit is used to receive edge node data from the edge node in the laboratory. The edge node is connected to multiple sensors installed in the laboratory and on multiple smart devices. The multiple sensors are used to collect monitoring data from the multiple smart devices and experimental environment data from the laboratory. The edge node data is obtained by the edge node after preprocessing the monitoring data and the experimental environment data. The generation unit is used to generate health operation thresholds corresponding to the multiple smart devices based on the historical operating data of the multiple smart devices, the environmental baseline of the laboratory, and the fault records of the multiple smart devices. The risk assessment unit is used to perform weighted scoring on the edge node data using a random forest algorithm, and to classify the risk levels of the multiple smart devices according to the weighted scores and the healthy operation threshold. The risk level characterizes the degree of abnormality in the operation of the smart devices. The greater the deviation between the edge node data and the healthy operation threshold, the higher the degree of abnormality. The risk levels include: Level 1 risk reflecting low degree of abnormality, Level 2 risk reflecting relatively high degree of abnormality, and Level 3 risk reflecting high degree of abnormality. The early warning unit is used to execute early warning strategies according to the risk level classification in order to eliminate malfunction events of the multiple smart devices. The adjustment unit is used to establish the correlation between the fault events of the multiple smart devices and the experimental batch information and changes in the experimental environment, and dynamically adjust the health operation threshold according to the correlation.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory being used to store one or more programs and configured to be executed by the processor, the programs including instructions for performing the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program for storing electronic data interchange, wherein the computer program causes a computer to perform the method as described in any one of claims 1-7.

Images