A monitoring and early warning method and system based on laboratory operation

Abstract

The application discloses a kind of monitoring early warning method and system based on laboratory operation, comprising, laboratory equipment state data and personnel operation record are acquired and handled, dynamic asymmetric bayesian network reflecting the dynamic relationship between personnel operation intention, equipment state and interference is constructed;Output from equipment state to personnel operation intention is deduced Probability distribution, monitor whether equipment state is abnormal, update probability distribution;Based on ROC curve, the updated probability distribution is judged and early warned, and the dynamic asymmetric bayesian network is defined Loss function dynamically adjusts;The application constructs dynamic asymmetric bayesian network, dynamically adjusts parameters according to the real-time change of equipment state, so that the model shows stronger robustness in the dynamically changing laboratory environment;By establishing the reverse inference mechanism of dynamic asymmetric bayesian network, personnel operation intention is inferred from equipment state, the perception level of potential risk is improved, and the deficiency of traditional system in deep inference is made up.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laboratory monitoring and early warning, and in particular to a monitoring and early warning method and system based on laboratory operation. BACKGROUND

[0002] Laboratory safety monitoring and early warning technology, as an indispensable safeguard means in modern scientific research environment, has made great progress in technology and application in recent years. Early laboratory monitoring mainly relies on video monitoring equipment, which realizes preliminary safety state evaluation and early warning function by monitoring personnel activity image and equipment running state. With the popularity of Internet of Things (IoT) technology, remote state monitoring and management in laboratory environment gradually emerges, making early identification and preventive maintenance of equipment running abnormity possible. In addition, the introduction of artificial intelligence and machine learning technology further promotes technological innovation, such as classifying personnel behavior patterns through deep learning algorithm or predicting potential risks by using time series data analysis. However, these technologies often have limitations in realizing correlation analysis of laboratory environment, especially in monitoring the interaction between personnel operation intention, equipment state and interference.

[0003] Among them, it is particularly obvious in the processing of high complexity and high risk scenarios. Traditional monitoring systems are difficult to adapt to the variability of personnel behavior and the dynamic evolution of equipment state in laboratory operation, and cannot accurately obtain the equipment state, personnel intention and interference relationship in the experiment process. In addition, the existing technology lacks in the integration and analysis ability of multi-source heterogeneous data, and fails to fully explore the deep correlation between personnel operation, equipment running state and hidden interference factors (such as someone trying to deliberately block the monitoring equipment), which limits the comprehensive perception and accurate early warning of potential risks by the system. SUMMARY

[0004] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, abstract and title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0005] In view of the above existing problems, the present application is proposed. Therefore, the present application provides a monitoring and early warning method based on laboratory operation to solve the problems mentioned in the background art.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a monitoring and early warning method based on laboratory operation, comprising:

[0008] Through the sensor network and the laboratory management system, laboratory equipment state data and personnel operation records are acquired and processed, and a dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, equipment state and interference is constructed;

[0009] The dynamic asymmetric Bayesian network is used to output a probability distribution from the equipment state to the personnel operation intention, and to monitor whether the equipment state is abnormal, and the probability distribution is updated according to the equipment state monitoring result;

[0010] The updated probability distribution is judged based on a ROC curve, a warning is given according to the judgment result, and a loss function is defined to dynamically adjust the dynamic asymmetric Bayesian network, so that intelligent monitoring and warning of user violation behavior under laboratory operation are realized.

[0011] As a preferred scheme of the monitoring and warning method under laboratory operation, the dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, equipment state and interference is constructed, including:

[0012] The nodes in the dynamic Bayesian network are defined as personnel operation intention O, equipment state S and interference relationship H, respectively;

[0013] The personnel operation intention O takes values of compliant operation and violation operation, the equipment state S takes values of normal operation and abnormal operation, and the interference relationship H takes values of no interference state and interference state;

[0014] The dynamic Bayesian network is defined as an asymmetric form, wherein the edge relationship of the nodes is O→S, representing the direct influence of personnel operation intention on equipment state; H→S, representing the direct influence of interference relationship on equipment state; and no O→H edge, representing that personnel operation intention does not directly affect the interference relationship.

[0015] As a preferred scheme of the monitoring and warning method under laboratory operation, the dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, equipment state and interference is constructed, including:

[0016] The weight of H→S is dynamically adjusted according to the switching frequency of the equipment state, and the formula is:

[0017] w H→S (t)=α·Var(S t-k:t )+(1-α)·w H→S (t-1)

[0018] Wherein, w H→S (t) represents the edge weight from the interference relationship H to the equipment state S at time t; α represents a smoothing factor for balancing the importance of the variance at time t-k and the edge weight at time t-1; Var(S t-k:t) represents the variance of the device state S from time t-k to time t; w H→S (t-1) represents the edge weight from H to S at time t-1.

[0019] As a preferred scheme of the monitoring and early warning method based on laboratory operation, wherein: the dynamic asymmetric Bayesian network is used to output the probability distribution from the device state to the personnel operation intention, including:

[0020] According to the personnel operation intention O and the interference relationship H, the probability distribution of the device state S is calculated, and a pseudo-likelihood function is constructed by setting the prior probability of the personnel operation intention;

[0021] The posterior probability of the interference relationship H is calculated, which is calculated by marginalizing the posterior probability of the personnel operation intention O;

[0022] According to the pseudo-likelihood function, the posterior probability of the interference relationship H, and the probability distribution property rule, the probability distribution from the device state to the personnel operation intention, i.e. the posterior probability, is output by iteration.

[0023] As a preferred scheme of the monitoring and early warning method based on laboratory operation, wherein: the iteration method includes:

[0024] Starting from the initial probability distribution of the set prior probability of the personnel operation intention, i.e. taking the set prior probability of the personnel operation intention as the iteration starting point;

[0025] In each iteration, the pseudo-likelihood function and the posterior probability of the interference relationship H are used to calculate the new posterior probability, i.e. the posterior probability from the device state to the personnel operation intention, and the posterior probability of the interference relationship H is updated through the personnel operation intention O;

[0026] With the increase of the number of iterations, the posterior probability from the device state to the personnel operation intention tends to the prior probability of the personnel operation intention, until the absolute value of the probability change between the previous and the next iteration is less than 0.01, and the iteration is stopped.

[0027] As a preferred scheme of the monitoring and early warning method based on laboratory operation, wherein: the device state is monitored for abnormality, and the probability distribution is updated according to the device state monitoring result, including:

[0028] The device state monitoring result is obtained by counting the number of times the device state switches from normal to abnormal and the maximum value of the duration of each abnormal state in seconds;

[0029] If it is determined that the equipment state monitoring result is abnormal, then according to the abnormal value of the equipment state monitoring result, the posterior probability of the personnel operation intention deduced from the equipment state is updated.

[0030] As a preferred solution of the monitoring and early warning method under laboratory operation, wherein: the updated probability distribution is judged based on the ROC curve, the early warning is performed according to the judgment result, and the loss function is defined to dynamically adjust the dynamic asymmetric Bayesian network, including:

[0031] The ROC curve is drawn through the updated probability distribution, if the posterior probability of the personnel operation intention deduced from the equipment state after updating is greater than the optimal threshold value in the generated ROC curve, the local sound and light early warning is triggered, and a short message is sent for notification, the early warning record and the short message notification record are uploaded to the laboratory management system;

[0032] At the same time, the loss function is defined according to the non-alarm condition, and the edge weight in the dynamic asymmetric Bayesian network is subjected to gradient descent.

[0033] In the second aspect, the present application provides a monitoring and early warning system based on laboratory operation, which comprises:

[0034] The dynamic asymmetric Bayesian network construction module is configured to acquire and process laboratory equipment state data and personnel operation records through a sensor network and a laboratory management system, and construct a dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, equipment state and interference;

[0035] The equipment state monitoring module is configured to output the probability distribution of the personnel operation intention deduced from the equipment state by using the dynamic asymmetric Bayesian network, monitor whether the equipment state is abnormal, and update the probability distribution according to the equipment state monitoring result;

[0036] The early warning and dynamic asymmetric Bayesian network dynamic adjustment module is configured to judge the updated probability distribution based on the ROC curve, perform early warning according to the judgment result, and define the loss function to dynamically adjust the dynamic asymmetric Bayesian network, so as to realize intelligent monitoring and early warning under laboratory operation.

[0037] In the third aspect, the present application provides a computer device comprising a memory and a processor, the memory stores a computer program, wherein: the processor implements any step of the above method when executing the computer program.

[0038] In the fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein: the computer program is executed by a processor to implement any step of the above method.

[0039] Compared with the prior art, the application has the beneficial effects that:

[0040] 1、The application constructs a dynamic asymmetric Bayesian network through personnel operation intention, equipment state and hidden interference, the network model can dynamically adjust parameters according to real-time changes of the equipment state, accurately captures the interaction between operation and state, improves the detection accuracy of abnormal conditions in the laboratory, and makes the model show stronger robustness in the dynamically changing laboratory environment; in addition, by establishing a reverse inference mechanism in the dynamic asymmetric Bayesian network, the personnel operation intention is inferred from the equipment state, and the subtle influence of hidden interference is identified through iterative optimization, the perception and early warning level of potential risks are improved, and the shortcomings of traditional systems in deep inference are made up;

[0041] 2、The sensor network and the laboratory management system are used to collect equipment state data and personnel operation records, and fuse them into the dynamic asymmetric Bayesian network for analysis, which provides a comprehensive view for the actual laboratory operation environment, so that the laboratory management system can timely and accurately issue early warning, and effectively reduces the possibility of false alarm caused by data fragmentation. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0043] Figure 1 The overall flowchart of the monitoring and early warning method based on laboratory operation according to an embodiment of the application;

[0044] Figure 2 The iteration flowchart of the probability distribution from equipment state back to personnel operation intention of the monitoring and early warning method based on laboratory operation according to an embodiment of the application;

[0045] Figure 3 The comparison chart of the influence of dynamic weight adjustment on warning accuracy of the monitoring and early warning method based on laboratory operation according to an embodiment of the application;

[0046] Figure 4 The ROC curve and iteration convergence process (personnel operation detection scene) chart of the monitoring and early warning method based on laboratory operation according to an embodiment of the application. DETAILED DESCRIPTION

[0047] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the protection scope of the present application.

[0048] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. The present application, however, can be practiced in a variety of ways beyond the specific details set forth herein without departing from the scope of the present application. It can be appreciated by those skilled in the art that the present application can be practiced without such specific details.

[0049] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. The "in one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or selected from other embodiments.

[0050] The present application is described in detail in conjunction with the schematic drawings. In the detailed description of the embodiments of the present application, the cross-sectional view of the device structure is partially enlarged without the general proportion for the convenience of description, and the schematic drawings are only examples, which should not limit the scope of protection of the present application. In addition, the three-dimensional spatial dimensions of length, width and depth should be included in actual manufacture.

[0051] Meanwhile, in the description of the present application, it should be noted that the terms "upper, lower, inner and outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first, second or third" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0052] Unless otherwise specifically defined and limited, the terms "mounting, connecting, connection" in the present application should be understood broadly, for example: it can be fixed connection, detachable connection or integral connection; it can also be mechanical connection, electrical connection or direct connection, it can also be indirectly connected through intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0053] Example 1

[0054] Reference Figure 1 and Figure 2For the first embodiment of the present application, the embodiment provides a monitoring and early warning method based on laboratory operation, comprising:

[0055] S1, through the sensor network and the laboratory management system, the laboratory equipment state data and personnel operation record are acquired and processed, and the dynamic asymmetric Bayesian network reflecting the personnel operation intention, the equipment state and the interference relationship is constructed;

[0056] Specifically, the laboratory equipment state data is acquired in real time through the sensor network, and the data acquisition includes temperature, pressure, current and the like; and the collected data is received through the stream processing framework Apache Kafka, a data packet is generated every second, the format is JSON, and it contains a timestamp and a state value, for example, {"timestamp":"2025-03-03 10:00:01","temperature":25.3,"pressure":101.2};

[0057] It should be noted that when the equipment state data is acquired, the equipment state data needs to be denoised, wherein the denoising adopts a sliding average filtering formula, and the purpose is to align the time stamp of the device data removed from the disturbance with the data in the laboratory management system;

[0058] Specifically, the operation behavior (device switching time, device parameter adjustment record, current laboratory user and corresponding personnel violation record, etc.) is recorded through the laboratory management system, and the time stamp (error less than 0.1 second) is aligned with the equipment state data;

[0059] Further, the nodes in the dynamic Bayesian network are defined as personnel operation intention O, equipment state S and interference relationship H respectively;

[0060] It should be explained that the traditional Bayesian network is a graphical model for representing random variables and their probability dependency, wherein the node represents the random variable, the edge represents the causal or dependent relationship between variables, and the direction of the edge usually reflects the direction of causality; in addition, the edge relationship in the traditional Bayesian network is bidirectional or symmetric, while in the asymmetric form Bayesian network constructed in the present application, the direction of the edge is unidirectional, and the edge relationship is not completely equal;

[0061] Specifically, the personnel operation intention O takes the value of compliant operation and non-compliant operation, the equipment state S takes the value of normal operation and abnormal operation, and the interference relationship H takes the value of no interference state and interference state;

[0062] Further, the dynamic Bayesian network is defined in an asymmetric form, where the edge relationship of the nodes is O→S, representing the direct influence of the personnel operation intention on the device state (meaning that the personnel operation intention will directly cause the device state to change regardless of compliance or violation; for example, a compliant operation can keep the device normal, while a violation operation can cause the device to be abnormal); H→S, representing the direct influence of the interference relationship on the device state (for example, shielding the camera or tampering with the data, which will affect the observed state of the device, and if the device is in an abnormal state, it will be masked as a normal state); and no O→H edge, representing that the personnel operation intention does not directly affect the interference relationship (indicating that the operation intention and the interference relationship are independent variables, in other words, the existence of interference is not directly determined by the personnel operation intention, but occurs independently).

[0063] Specifically, the values of the personnel operation intention O, the device state S, and the interference relationship H and the edge relationship therebetween are explained; for the edge relationship O→S, simply speaking, if the personnel operation intention is compliant (the experimental personnel operates the instrument according to the regulations), the device state will generally remain normal, because the compliant operation will not damage the device; if the personnel operation intention is a violation, it can cause the device state to become abnormal (the experimental personnel secretly adjusts the instrument parameters); but this is only the general case, because the device state S is also affected by the interference relationship H; for H→S, the interference can also affect the device state, where the interference can be intentional behavior (such as shielding the monitoring camera), or it can be an accidental situation (such as a device failure due to its own reasons); then without interference, the device state can directly reflect the result of the personnel operation intention (the experimental personnel violates the operation, causing the device to be abnormal); and with interference, the interference can mask or distort the device state; (for example, the experimental personnel violates the operation, causing the device to be abnormal, but if he also shields the monitoring camera, the monitoring camera can observe that the device state is normal, because the device abnormality is hidden); for no O→H, it is indicated that the occurrence of the interference does not depend on the personnel operation intention, that is, when the personnel operation intention is compliant, there can be interference or no interference; when the personnel operation intention is a violation, there can also be interference or no interference; in addition, if there is O→H, it indicates that the personnel operation intention directly determines the interference, that is, a violation intention always causes interference, and a compliant intention always has no interference; but the reality is not like this, for example, a personnel operation intention that is compliant can accidentally shield the camera, and a personnel operation intention that is a violation can not take any interference measures (shield the monitoring camera);

[0064] It should be noted that, due to the laboratory environment is not constant, the equipment state may be affected by many factors, such as equipment aging, external interference or change of operation habit, if the fixed edge weight is used, the dynamic asymmetric Bayesian network cannot timely reflect these changes, resulting in the decrease of prediction or detection accuracy; then if the weight of H→S is dynamically adjusted, the model can update the model according to the real-time equipment state data, so that the edge weight is more in line with the current laboratory environment;

[0065] Further, according to the switching frequency of the equipment state, the weight of H→S is dynamically adjusted, and the formula is represented as:

[0066] w H→S (t)=α·Var(S t-k:t )+(1-α)·w H→S (t-1)

[0067] Wherein, w H→S (t) is represented as the edge weight from the interference relationship H to the equipment state S at time t;Alpha is represented as a smoothing factor, which is used to balance the importance of the variance at time t-k and the edge weight at time t-1;Var(S t-k:t ) represents the variance of the equipment state S from time t-k to time t;w H→S (t-1) represents the edge weight from H to S at time t-1;

[0068] It should be noted that, if the switching frequency of the equipment state is high, it means that the influence of the interference on the equipment state is enhanced, at this time, by increasing the weight of H→S, the sensitivity of the model to the abnormality can be improved;When the switching frequency of the equipment state is low, it means that the influence of the interference on the equipment state is weak, and the weight adjustment amplitude can be appropriately reduced, so that the model can quickly respond to the equipment state;

[0069] S2, using the dynamic asymmetric Bayesian network, outputting the probability distribution from the equipment state to the personnel operation intention, and monitoring whether the equipment state is abnormal, according to the equipment state monitoring result, updating the probability distribution;

[0070] It should be noted that, due to the asymmetric of the Bayesian network structure and the one-way introduction of the interference relationship H, the traditional Bayesian inference method is difficult to be directly applied, and the method scheme of the present application proposes a probability reconstruction mechanism based on pseudo-likelihood function and iterative updating, so as to realize the reverse inference from the equipment state to the personnel operation intention;

[0071] Further, according to the personnel operation intention O and the interference relationship H, the probability distribution of the equipment state S is calculated, and the pseudo-likelihood function is constructed by setting the prior probability of the personnel operation intention;

[0072] Specifically, the pseudo-likelihood function L(O|S, H) can be expressed by a mathematical formula as follows:

[0073] L(O|S, H) = P(S|O, H) · P(O)

[0074] wherein P(O) represents the prior probability of the set personnel operation intention, and the initial value set is subject to a uniform distribution (for example, the probabilities of the compliant personnel operation intention and the non-compliant personnel operation intention are equal); P(S|O, H) represents the probability distribution of the obtained device state S under the joint action of the personnel operation intention O and the interference relationship H; assuming that O takes the value of compliance and H takes the value of no interference, the result obtained is the probability distribution of the device state under the personnel operation intention compliance and no interference;

[0075] Further, the posterior probability of the interference relationship H is calculated by marginalizing the posterior probability of the personnel operation intention O;

[0076] Specifically, the posterior probability of the personnel operation intention O is marginalized by the posterior probability of the operation intention O:

[0077]

[0078] wherein P (k) (H|S) represents the posterior probability of the interference relationship H in the kth iteration; P (k) (O|S) represents the posterior probability of the personnel operation intention O in the kth iteration;

[0079] Further, according to the pseudo-likelihood function, the posterior probability of the interference relationship H, and the probability distribution property rule, the probability distribution from the device state to the personnel operation intention, that is, the posterior probability, is output by iteration.

[0080] Specifically, the posterior probability from the device state to the personnel operation intention is output as follows:

[0081]

[0082] wherein ∑ O L(O|S, H) · P (k) (H|S) is a normalization factor, which ensures that P (k+1) (O|S) satisfies the probability distribution property rule, that is, the sum is 1.

[0083] Specifically, the posterior probability from the device state to the personnel operation intention is output as follows: Figure 2 wherein the steps of the iteration are as follows:

[0084] S201, starting from the initial probability distribution as the prior probability of the set personnel operation intention, that is, taking the prior probability of the set personnel operation intention as the iteration starting point: P (0)(O|S) = P(O);

[0085] S202、In each iteration, the new posterior probability P (k+1) (O|S), that is, the posterior probability of inferring the personnel operation intention from the device state, and updating the posterior probability P (k) (H|S);

[0086] S203、With the increase of the number of iterations, the posterior probability of inferring the personnel operation intention from the device state tends to the prior probability of the personnel operation intention, until the absolute value of the probability change between the previous and the next iteration is less than 0.01, that is, |P (k+1) (O|S)-P (k) (O|S)|<0.01, stop iteration;

[0087] It should be noted that through the iterative operation, P(O|S) is gradually optimized, avoiding directly solving the normalization constant P(S) in the Bayesian network, and overcoming the limitations of the traditional Bayesian inference method in the asymmetric network and the interference relationship scenario;

[0088] Further, by counting the number of times the device state switches from normal to abnormal and the maximum value of the duration of each abnormal state in seconds, the device state monitoring result is obtained;

[0089] Specifically, within a time window of 60s, the number of times the device state switches from normal to abnormal is counted, the switching condition is defined as the specific device state parameter exceeding the normal range (such as temperature > 30℃), and the duration of each abnormal state in seconds is recorded, and the maximum value is taken;

[0090] It should be noted that if the sensor data transmission is interrupted, the device state is marked as missing, the calculation process is suspended, and a warning is issued, and the device state monitoring result is displayed as abnormal;

[0091] Further, if it is determined that the device state monitoring result is abnormal, the posterior probability of inferring the personnel operation intention from the device state is updated according to the abnormal value of the device state monitoring result;

[0092] Specifically, the posterior probability P (k +1) (O|S) is updated by using a Sigmoid function to perform smooth adjustment, that is, Where ΔF is the abnormal value of the device state monitoring result;

[0093] S3, judging the updated probability distribution based on the ROC curve, warning according to the judgment result, and defining the loss function to dynamically adjust the dynamic asymmetric Bayesian network, so as to realize intelligent monitoring and warning of user violation behavior under laboratory operation;

[0094] Further, the ROC curve is drawn based on the updated probability distribution, if the posterior probability from the device state to the personnel operation intention after updating is greater than the optimal threshold in the generated ROC curve, the local sound and light warning is triggered, a short message is sent for notification, the warning record and the short message notification record are uploaded to the laboratory management system;

[0095] Specifically, the updated P (k+1) (O|S), draw the ROC curve, wherein each point on the ROC curve corresponds to a specific theta value, and each point on the ROC curve is obtained by taking (FPR, TPR) as coordinates, that is, false positive rate and true positive rate, select the theta that makes FPR and TPR maximum as the optimal threshold through Youden index;

[0096] Specifically, the local sound and light warning displays the laboratory light in red flashing state, sends a short message and broadcasts, and uploads the warning record and the short message notification record to the laboratory management system; The format is: warning: warning language + warning timestamp; For example, "warning: laboratory equipment is abnormal, time: 2025-03-03 10:00:01";

[0097] It should be noted that by selecting the optimal threshold on the ROC curve, the model can ensure a high true positive rate when issuing a warning, while controlling the false positive rate within an acceptable range; On the one hand, it can improve the reliability of the warning, and on the other hand, it can avoid the problem of wasting resources due to false warning;

[0098] Further, the loss function is defined according to the non-warning condition, and the edge weight in the dynamic asymmetric Bayesian network is subjected to gradient descent;

[0099] It should be noted that if the non-warning condition is met, the edge weight in the dynamic asymmetric Bayesian network needs to be subjected to gradient descent to ensure that the model still maintains high reasoning accuracy in the idle state;

[0100] Specifically, the loss function L is defined as:

[0101] L = -∑logP(S|O,H,w H→S )

[0102] Specifically, the gradient descent formula is represented as:

[0103]

[0104] wherein, is updated w H→S , is pre-updated w H→S , and η is a learning rate.

[0105] Further, the embodiment also provides a monitoring and early warning system based on laboratory operation, comprising:

[0106] A dynamic asymmetric Bayesian network construction module is configured to acquire and process laboratory equipment state data and personnel operation records through a sensor network and a laboratory management system, and construct a dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, equipment state and interference;

[0107] An equipment state monitoring module is configured to output a probability distribution from equipment state back to personnel operation intention by using the dynamic asymmetric Bayesian network, monitor whether the equipment state is abnormal, and update the probability distribution according to the equipment state monitoring result;

[0108] An early warning and dynamic asymmetric Bayesian network dynamic adjustment module is configured to judge the updated probability distribution based on a ROC curve, perform early warning according to the judgment result, and dynamically adjust the dynamic asymmetric Bayesian network by defining a loss function, so as to realize intelligent monitoring and early warning under laboratory operation.

[0109] The embodiment also provides a computer device suitable for the monitoring and early warning method based on laboratory operation, comprising:

[0110] A memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the monitoring and early warning method based on laboratory operation proposed in the above embodiment.

[0111] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0112] The embodiment also provides a storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the monitoring and early warning method based on laboratory operation according to the above embodiment.

[0113] The storage medium according to the embodiment and the data storage method according to the above embodiment belong to the same inventive concept, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0114] Embodiment 2

[0115] Reference Figure 3 and Figure 4 As a second embodiment of the present application, the embodiment provides a monitoring and early warning method based on laboratory operation, which includes: verifying the beneficial effects in the scheme of the present application through experiments; selecting a chemical laboratory of a university as a test scene, and the test objects include a high-pressure reaction kettle (model: Parr 4560), a gas chromatograph (model: Agilent 7890B) and personnel operation behaviors; a sensor network is composed of the following devices: a temperature sensor: Omega HH314 (accuracy ±0.1℃, sampling frequency 1Hz); a pressure sensor: Honeywell TruStability HSC series (accuracy ±0.5% FS, range 0-10MPa); a current monitoring module: NI cDAQ-9188 (accuracy ±0.01A); a camera: FLIR A315 thermal imaging camera (resolution 640×480, frame rate 30fps);

[0116] The prior art control group adopts: fixed threshold alarm and traditional Bayesian network; wherein, the fixed threshold alarm is a static threshold of setting device parameters (such as temperature > 80℃ triggers alarm); wherein, the traditional Bayesian network is a Bayesian network based on fixed edge weight (O→S=0.7, H→S=0.3);

[0117] The experimental results are shown as Figure 3 and Figure 4 It can be seen from Figure 3 that the dynamic adjustment method of the scheme can still maintain a higher accuracy rate when the sample increases, verifying its advantage in the laboratory dynamic environment; secondly, from Figure 4 (left figure) it can be seen that the AUC of the scheme is about 0.97, which is significantly higher than the AUC of the traditional method about 0.82; at the same time, through Figure 4 (right figure) it can be found that the highest value 1 of the posterior probability is reached after 5 iterations, indicating that the stability of the posterior probability can be realized through a short number of iterations;

[0118] In summary, the network model established by the scheme can dynamically adjust the parameters according to the real-time changes of the device state when the sample increases, and the stability of the model can be realized through a short number of iterations, which provides technical support for monitoring and early warning in the real laboratory operating environment.

[0119] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code. The schemes in the embodiments of the present application can be implemented in various computer languages, such as object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0120] The present application is described with reference to flowcharts and / or block diagrams according to the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a machine that implements the flowcharts and / or block diagrams. Figure 1 one flow or multiple flows and / or blocks Figure 1apparatus for performing each function specified in a flow or flows and / or blocks

[0121] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions Figure 1 a flow or flows and / or blocks Figure 1 an apparatus for performing each function specified in a flow or flows and / or blocks

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions Figure 1 a flow or flows and / or blocks Figure 1 an apparatus for performing each function specified in a flow or flows and / or blocks

[0123] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the preferred embodiments by those of skill in the art once they have the benefit of the present disclosure. Therefore, the appended claims are intended to encompass within their scope all possible variations and modifications of the preferred embodiments.

[0124] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A monitoring and early warning method based on laboratory operation, characterized in that, The method comprises the following steps: Through the sensor network and the laboratory management system, laboratory equipment state data and personnel operation records are acquired and processed, and a dynamic asymmetric Bayesian network reflecting the relationship among personnel operation intention, equipment state and interference is constructed; The construction of the dynamic asymmetric Bayesian network reflecting the relationship among personnel operation intention, equipment state and interference comprises the following steps: Nodes in the dynamic Bayesian network are defined as personnel operation intention O, equipment state S and interference H respectively; The personnel operation intention O takes values of compliant operation and non-compliant operation, the equipment state S takes values of normal operation and abnormal operation, and the interference H takes values of non-interference state and interference state; The dynamic Bayesian network is defined as an asymmetric form, wherein the edge relationship of the nodes is O→S, representing the direct influence of personnel operation intention on equipment state; H→S, representing the direct influence of interference on equipment state; and no O→H edge, representing that personnel operation intention does not directly affect interference; The method further comprises the following steps: According to the switching frequency of the equipment state, the weight of H→S is dynamically adjusted, and the formula is: wherein, represents an edge weight from an interference relation H to a device state S at time t; represents a smoothing factor balancing the importance of the variance at time t-k and the edge weight at time t-1 ; represents a variance of a device state S from time t-k to time t; represents an edge weight from H to S at time t-1 ; Using the dynamic asymmetric Bayesian network, the probability distribution from the equipment state to the personnel operation intention is output, and whether the equipment state is abnormal is monitored; according to the equipment state monitoring result, the probability distribution is updated; Based on the ROC curve, the updated probability distribution is judged, a warning is given according to the judgment result, and a loss function is defined to dynamically adjust the dynamic asymmetric Bayesian network, so as to realize intelligent monitoring and warning of user non-compliant behavior in laboratory operation.

2. The monitoring and warning method based on laboratory operation according to claim 1, wherein, Using the dynamic asymmetric Bayesian network, the probability distribution from the equipment state to the personnel operation intention comprises the following steps: According to the personnel operation intention O and the interference H, the probability distribution of the equipment state S is calculated, and a pseudo-likelihood function is constructed by setting the prior probability of the personnel operation intention; The posterior probability of the interference H is calculated, and the posterior probability of the interference H is calculated by marginalizing the posterior probability of the personnel operation intention O; According to the pseudo-likelihood function, the posterior probability of the interference H and the probability distribution property rule, the probability distribution from the equipment state to the personnel operation intention, i.e. the posterior probability, is output by iteration.

3. The monitoring and warning method based on laboratory operation according to claim 2, wherein, The iteration method comprises the following steps: Starting from the initial probability distribution, i.e. the prior probability of the personnel operation intention set as the iteration starting point; In each iteration, the pseudo-likelihood function and the posterior probability of the interference H are used to calculate the new posterior probability, i.e. the posterior probability from the equipment state to the personnel operation intention, and the posterior probability of the interference H is updated through the personnel operation intention O; With the increase of the number of iterations, the posterior probability from the equipment state to the personnel operation intention tends to the prior probability of the personnel operation intention, and the iteration is stopped until the absolute value of the probability change between the previous and the next iteration is less than 0.

01.

4. The monitoring and warning method based on laboratory operation according to claim 2, wherein, Monitoring whether the equipment state is abnormal, and updating the probability distribution according to the equipment state monitoring result comprises the following steps: The device state monitoring result is obtained by counting the number of times that the device state switches from normal to abnormal and the maximum value of the duration of each abnormal state in seconds; If it is determined that the device state monitoring result is abnormal, the posterior probability of the personnel operation intention inferred from the device state is updated according to the abnormal value of the device state monitoring result.

5. The monitoring and warning method based on laboratory operation according to claim 1 or 4, characterized in that, Based on the ROC curve, the updated probability distribution is judged, the warning is given according to the judgment result, and the loss function is defined to dynamically adjust the dynamic asymmetric Bayesian network, including: The ROC curve is drawn through the updated probability distribution, and if the posterior probability of the personnel operation intention inferred from the device state after updating is greater than the optimal threshold value in the generated ROC curve, a local sound-light warning is triggered, a short message is sent for notification, the warning record and the short message notification record are uploaded to the laboratory management system; At the same time, the loss function is defined according to the non-warning condition, and the edge weight in the dynamic asymmetric Bayesian network is subjected to gradient descent.

6. A monitoring and warning system under laboratory operation based on any one of claims 1 to 5, characterized in that, Including: The dynamic asymmetric Bayesian network construction module is configured to acquire and process laboratory device state data and personnel operation records through a sensor network and a laboratory management system, and construct a dynamic asymmetric Bayesian network reflecting the relationship between personnel operation intention, device state and interference; The device state monitoring module is configured to output the probability distribution of the personnel operation intention inferred from the device state by using the dynamic asymmetric Bayesian network, and monitor whether the device state is abnormal, and update the probability distribution according to the device state monitoring result; The warning and dynamic asymmetric Bayesian network dynamic adjustment module is configured to judge the updated probability distribution based on the ROC curve, give a warning according to the judgment result, and define a loss function to dynamically adjust the dynamic asymmetric Bayesian network, so as to realize intelligent monitoring and warning under laboratory operation. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the method of any one of claims 1-5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the method of any one of claims 1-5.

Images