Teaching resource dynamic management system and method based on digital intelligence interaction

Abstract

The invention discloses a teaching resource dynamic management system and method based on digital intelligence interaction, and relates to the technical field of teaching safety management, and the system comprises the steps: sensing multi-modal data, presetting a causal chain network of experiment operation, and constructing a teaching experiment risk map based on the causal chain network; acquiring multi-modal data in real time, generating a feature vector, and performing causal chain risk conduction calculation by combining the teaching experiment risk map with space-time attention; performing GAN risk scene simulation on the basis of the multi-modal data acquired in real time; and calculating a risk gain according to a GAN risk scene simulation result, and performing graded early warning based on causal chain risk conduction calculation and the risk gain. According to the method, closed-loop management from data acquisition, risk modeling to prediction and early warning is realized, and the teaching quality and the safety protection function in the experiment process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of teaching safety management, and specifically to a dynamic management system and method for teaching resources based on digital and intelligent interaction. Background Art

[0002] In chemistry, biology, and physics labs, how can we leverage sensors (motion capture, environmental monitoring) and AI vision to analyze students' procedures for compliance and potential risks in real time? When high-risk actions (such as incorrect chemical mixing or incorrectly performing procedures) or environmental anomalies (such as the risk of toxic gas leaks) are detected, how can the system issue warnings to prevent the occurrence of danger? How can we promptly transmit warning signals to the experimenter or other relevant personnel? This is an urgent problem that needs to be solved. Summary of the Invention

[0003] The purpose of the present invention is to provide a system and method for dynamic management of teaching resources based on digital and intelligent interaction to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic management of teaching resources based on digital and intelligent interaction, the method comprising: Step S100: sensing multimodal data, where the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; Step S200: Preset a causal chain network of experimental operations, and construct a teaching experiment risk map based on the causal chain network; Step S300: Acquire multimodal data in real time, generate feature vectors, combine the teaching experiment risk map, and integrate spatiotemporal attention to perform causal chain risk conduction calculation; Step S400: Perform a GAN risk scenario simulation based on the multimodal data acquired in real time; calculate the risk gain based on the GAN risk scenario simulation results, and perform a graded warning based on the causal chain risk transmission calculation and risk gain.

[0005] Furthermore, step S100 specifically includes: Perception of multimodal data, the multimodal data including at least one of instantaneous image snapshots, sensory data from experimental instruments, sensory data from experimenters, and environmental data; Install a camera on the teaching experiment table. Preferably, a high-frame-rate AI camera is selected to continuously take instantaneous image snapshots during the experiment. The instantaneous image snapshots include experimental operation actions, tool positions, and the addition process of chemical raw materials; Equip the experimental instruments with sensors, including pressure sensors, temperature sensors, and speed sensors, etc., to obtain sensor data of the experimental instruments such as pressure, instrument surface temperature, and instrument speed during the experiment; When necessary experiments are performed, the experimenter needs to wear smart experiment gloves equipped with pressure sensors, etc., to obtain the experimenter's sensory data such as hand pressure distribution and tremor frequency during the experiment; Install various sensors in the laboratory, including temperature sensors, humidity sensors, and gas concentration sensors, to collect laboratory environmental data; It should be noted that the high-risk points for dangers vary for different experiments. For chemical experiments, the order of adding various chemical raw materials, the concentrations of chemical reactants and products, and the experimental steps are more important. In particular, chemical reactants and products may be hazardous. When their gases reach a certain concentration, they will cause harm to the experimenters. Therefore, the focus is on instantaneous image snapshots and environmental data. Therefore, according to the actual situation, at least one of the instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data is selected for safety supervision to ensure the safety of personnel during the experiment. Furthermore, step S200 specifically includes: Preset a causal chain network of experimental operations, and construct a teaching experiment risk map based on the causal chain network; It should be noted that the causal chain network of the experimental operation can be preset by the experimental teaching teacher, and a large number of experimental teaching resources on the network can also be searched, but it must be verified by the experimental teaching teacher before it can be preset as the causal chain network of the experimental operation; The causal chain is defined as a five-tuple: C = [V, E, P, W, T]; where C represents the five-tuple identifier of an experiment; V represents the node set; E represents the directed edge set; P represents the transition probability matrix; W represents the result severity matrix; and T represents the time decay factor. The nodes include operations, risks and results; Including: V={Vi|i=1,2,…,n}, i represents the number of nodes, n represents the total number of nodes; The directed edges represent causal relationships, E⊆V×V; The transition probability matrix and the result severity matrix are obtained according to the causal chain network of the preset experimental operation; According to the five-tuple: C = [V, E, P, W, T], a teaching experiment risk map is constructed, specifically: The node set V in the five-tuple is used as the node of the teaching experiment risk map, and the directed edge set E in the five-tuple is used as the edge of the teaching experiment risk map; Each edge e u→vrepresents a directed edge from node u to node v, u, v∈V; each edge e u→v Associate a conduction weight w u→v , determined based on the ratio of all errors in the history from the operation steps corresponding to node u to the operation steps corresponding to node v, ranging from (0,1); Furthermore, in step S300, multimodal data is acquired in real time to generate feature vectors, and the causal chain risk conduction calculation is performed by integrating spatiotemporal attention with the teaching experiment risk map; Step S301: During the experiment, instantaneous image snapshots are acquired in real time, processed by a pre-trained 3D-CNN model, the operator's skeletal key points and tool / reagent positions are extracted, and the action deviation A is calculated based on the Hausdorff distance with the standard template; Through the experiment, the sensor equipped on the instrument is used to read the sensor data value of the experimental instrument in real time, and the abnormal value M is calculated based on the deviation of the preset safety range; Through the smart experimental gloves, the pressure distribution and tremor frequency of the experimenter's hand are obtained in real time, and the current experimenter's tension index B is calculated based on historical data; Through various sensors installed in the laboratory, sensor data values ​​are obtained in real time to calculate the weighted risk coefficient S of indicators such as gas concentration, temperature and humidity; A feature vector (A, M, B, S) is formed by using at least one of the action deviation A, the abnormal value M, the stress index B, and the risk factor S; Step S302: Calculate the inherent risk value R of each node based on the composed feature vector (A, M, B, S) base , characterized by: R base =αA+βM+γB+δS, where α, β, γ, and δ are weight coefficients, and normalization ensures R base ∈[0,1]; Combined with the teaching experiment risk map, the spatiotemporal attention is integrated to calculate the causal chain risk conduction; Furthermore, step S302 is specifically as follows: Step S3021: locating the corresponding node of the teaching experiment risk map according to the current operation step; Step S3022: recursively traverse all upstream operations of the node and calculate the precursor risk transmission amount; Starting from the current node, recursively operate upstream, and the formula is represented as: R cond (u)=R base (u)+Σ p∈parents(u) w p→u ×R cond (p) × T(p); Among them, R cond(u) represents the conduction risk value of node u; parents(u) represents the set of direct predecessor nodes of node u, and p represents the elements in the set of direct predecessor nodes; R base (u) represents the inherent risk of node u; w p→u represents the transmission weight from node p to node u; R cond (p) represents the conduction risk value of node p; T(p) represents the time attenuation factor of node p; The calculation formula of the precursor risk transmission amount is characterized as follows: P risk =Σ u∈parents(v) w u→v ×R cond (u)×T(u); where, P risk represents the predecessor conduction risk; parents(v) represents the set of direct predecessor nodes of node v; T(u) represents the time attenuation factor of node u; w u→v represents the transmission weight from node u to node v; When the operation interval time t is greater than the preset threshold, the -λt Calculate the time decay factor; when the operation interval time t is within the preset threshold, the time decay factor is always 1; Wherein, e represents the natural constant; t represents the operation interval time; λ represents the adjustment coefficient; It should be noted that the calculation formula for the precursor risk transmission amount aggregates the transmission risks of all direct predecessor nodes and multiplies them by time decay. The recursive process ensures that indirect predecessor risks are included, making the calculation of the precursor risk transmission amount more reasonable; If node u has no predecessor node (i.e. source node), then R cond (u)=R base (u); Step S3023: Based on the result severity matrix W, the severity of all possible results is accumulated, and the formula is: S=Σ k=1 K W v,k ; Where S represents the severity of the result, ranging from [0, K]; W v,k represents the severity value of the kth possible outcome of node v in the outcome severity matrix W, W v,k ∈[0,1]; K represents the number of possible outcomes of node v; It should be noted that the resulting severity matrix W is predefined, and each entry W v,k Represents a specific result. Since the severity matrix W is different for different experiments, the entry W can be obtained according to the number of experimental steps. v,k The number of , usually does not require a lot of calculation; Step S3024: Combine the precursor risk transmission amount and the current result severity to generate a transmission risk score of 0-1, specifically: Risk score =z×P risk +(1-z)×[S / S max ]; Among them, Risk score represents the conduction risk score; z represents the adjustment coefficient, which is used to coordinate the precursor risk conduction amount and the current result severity, z∈[0,1]; S max Indicates the maximum result severity of the node; Furthermore, in step S400, a GAN risk scenario simulation is performed based on the multimodal data acquired in real time; the risk gain is calculated based on the GAN risk scenario simulation results, and a graded warning is performed based on the causal chain risk transmission calculation and the risk gain; Step S401: pre-build a conditional GAN ​​model through deep learning, and input a noise vector and a real-time feature vector (A, M, B, S) into the conditional GAN ​​model generator; Inputting historical risk data into the conditional GAN ​​model discriminator; Step S402: Output the simulated feature vector (A) at the output end of the conditional GAN ​​model generator. sim ,M sim ,B sim ,S sim ); Step S403: derive the simulated feature vectors (A) from several groups of simulated outputs. sim ,M sim ,B sim ,S sim ) Quantify the risk increment of the simulated scenario compared to the current state, represented by the formula: Risk_Gain=Σ N max(0,Risk sim -Risk score ) / N; Among them, Risk_Gain represents the risk increment; N represents the number of GAN model groups; Risk sim represents the simulated conduction risk score calculated based on the simulated eigenvector; It should be noted that the risk increment is a relative increment, which only accumulates positive risks, ignores scenarios where the risk is reduced, and normalizes the results. Step S404: Trigger a graded warning based on the current conduction risk score and the simulated risk increment.

[0006] A dynamic teaching resource management system based on digital and intelligent interaction, including: data acquisition module, data retrieval module, calculation module, GAN simulation module and early warning module; The data acquisition module is used to acquire multimodal data in real time, wherein the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; The data retrieval module is used to preset a causal chain network of experimental operations, construct a teaching experiment risk map based on the causal chain network, store the teaching experiment risk map, and provide a retrieval function; The calculation module is responsible for calculation and realizes the automatic calculation process through the program; The GAN simulation module is used to pre-build a conditional GAN ​​model and simulate GAN risk scenarios based on multimodal data acquired in real time; The early warning module is used to quantify the risk increment of the simulated scenario compared with the current state through several groups of simulated feature vectors output by simulation, and trigger a graded early warning by combining the current conduction risk score and the simulated risk increment.

[0007] Specifically: the computing module supports independent computing.

[0008] Compared with the prior art, the beneficial effects of the present invention are: the present application realizes risk control of the entire process of teaching experiments, significantly improving safety and management efficiency. At the data collection level, cameras, sensors and other equipment are used to collect images, instrument status, personnel operations and environmental data, and differentiated monitoring is performed for different experimental types to ensure data accuracy. A risk map is constructed based on a preset causal chain network, and the risk transmission path is quantified with a five-tuple model, and the risk relationship is visualized. The spatiotemporal attention mechanism is integrated to calculate the inherent risk through the feature vector, and the transmission risk is recursively calculated in combination with the predecessor node, taking into account the influence of time and space dimensions. With the help of the conditional GAN ​​model, potential risk scenarios are simulated, the risk increment is quantified, and graded warnings are carried out in combination with real-time risk scores. This method realizes closed-loop management from data collection, risk modeling to prediction and warning, which can not only ensure experimental safety, but also dynamically optimize the allocation of teaching resources, improve teaching quality and safety protection functions during the experiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A schematic diagram of the steps of a method for dynamic management of teaching resources based on digital and intelligent interaction according to the present invention; Figure 2 This is a schematic diagram of the GAN risk scenario simulation process of the present invention; Figure 3 Schematic diagram of the structure of the conditional GAN ​​model of the present invention. DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0011] Example: Figure 1-Figure 3 As shown, the present invention provides a technical solution, a dynamic management method of teaching resources based on digital and intelligent interaction, the dynamic management method of teaching resources includes: Step S100: Perception of multimodal data, where the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; Step S200: Preset a causal chain network of experimental operations and construct a teaching experiment risk map based on the causal chain network; Step S300: Acquire multimodal data in real time, generate feature vectors, combine them with the teaching experiment risk map, and integrate spatiotemporal attention to perform causal chain risk conduction calculations; Step S400: Perform a GAN risk scenario simulation based on the multimodal data acquired in real time; calculate the risk gain based on the GAN risk scenario simulation results, and perform a graded warning based on the causal chain risk transmission calculation and risk gain.

[0012] Install a camera on the teaching lab bench. Preferably, choose a high-frame-rate AI camera. During the experiment, it continuously captures instantaneous image snapshots, including experimental operation actions, tool positions, and the addition process of chemical raw materials. Equip the experimental instruments with sensors, including pressure sensors, temperature sensors, and speed sensors, etc., to obtain sensor data of the experimental instruments such as pressure, instrument surface temperature, and instrument speed during the experiment; When necessary experiments are carried out, the experimenter needs to wear smart experiment gloves equipped with pressure sensors to obtain the experimenter's sensory data such as hand pressure distribution and tremor frequency during the experiment; Install various sensors in the laboratory, including temperature sensors, humidity sensors, and gas concentration sensors, to collect laboratory environmental data; It should be noted that the high-risk points for dangers vary for different experiments. For chemical experiments, the order of adding various chemical raw materials, the concentrations of chemical reactants and products, and the experimental steps are more important. In particular, chemical reactants and products may be hazardous. When their gases reach a certain concentration, they will cause harm to the experimenters. Therefore, the focus is on instantaneous image snapshots and environmental data. Therefore, according to the actual situation, at least one of the instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data is selected for safety supervision to ensure the safety of personnel during the experiment. Preset the causal chain network of experimental operations and construct the teaching experiment risk map based on the causal chain network; It should be noted that the causal chain network of the experimental operation can be preset by the experimental teaching teacher, and a large number of experimental teaching resources on the network can also be searched, but it must be verified by the experimental teaching teacher before it can be preset as the causal chain network of the experimental operation; The causal chain is defined as a five-tuple: C = [V, E, P, W, T]; where C represents the five-tuple identifier of an experiment; V represents the node set; E represents the directed edge set; P represents the transition probability matrix; W represents the result severity matrix; and T represents the time decay factor. Nodes include actions, risks, and outcomes; Including: V={Vi|i=1,2,…,n}, i represents the number of nodes, n represents the total number of nodes; Directed edges represent causal relationships, E⊆V×V; The transition probability matrix and outcome severity matrix are obtained based on the causal chain network of the pre-set experimental manipulations; According to the five-tuple: C = [V, E, P, W, T], a teaching experiment risk map is constructed, specifically: The node set V in the five-tuple is used as the node of the teaching experiment risk map, and the directed edge set E in the five-tuple is used as the edge of the teaching experiment risk map; Each edge e u→v represents a directed edge from node u to node v, u, v∈V; each edge e u→v Associate a conduction weight w u→v , determined based on the ratio of all errors in the history from the operation steps corresponding to node u to the operation steps corresponding to node v, ranging from (0,1); Furthermore, in step S300, multimodal data is acquired in real time to generate feature vectors, and the causal chain risk conduction calculation is performed by integrating spatiotemporal attention with the teaching experiment risk map; Step S301: During the experiment, instantaneous image snapshots are acquired in real time, processed by a pre-trained 3D-CNN model, the operator's skeletal key points and tool / reagent positions are extracted, and the action deviation A is calculated based on the Hausdorff distance with the standard template; The calculation method of Hausdorff distance is an existing technology. No collective calculation is performed here. The calculation results are normalized to obtain the action deviation A. Through the experiment, the sensor equipped on the instrument is used to read the sensor data value of the experimental instrument in real time, and the abnormal value M is calculated based on the deviation of the preset safety range; Among them, the preset safety range is obtained based on historical experience and is generally based on the historical average value; Through the smart experimental gloves, the pressure distribution and tremor frequency of the experimenter's hand are obtained in real time, and the current experimenter's tension index B is calculated based on historical data; The estimation of the stress index is a prior art. In this embodiment, a calculation method is provided: Real-time hand pressure distribution: f={f1,f2,...,fj} (pressure values ​​at j sensor points); Calculate the average pressure distribution f' based on historical benchmarks; The tension index B was estimated by pressure deviation; B=Σ j |(f-f') / max(f hist )| / j; where max(f hist ) The maximum pressure reference value in history is set manually based on experience; The tremor frequency and pressure distribution estimation methods are similar; Through various sensors installed in the laboratory, sensor data values ​​are obtained in real time to calculate the weighted risk coefficient S of indicators such as gas concentration, temperature and humidity;

[0013] D_Cg=min(1, Cg / Cg safe ); When the temperature WT is within the safety threshold range [WT min ,WT max ], the temperature risk coefficient D_WT is 0; when the temperature WT is within the safety threshold range [WT min ,WT max ], the temperature risk coefficient D_WT=min(1,|WT-WT mid | / (WT max -WT mid )); where WT mid =(WT max-WT min ) / 2; WT max Indicates the upper limit of the safety threshold; WT min Indicates the lower limit of the safety threshold; The same applies to the humidity risk factor; S=min(1,l1*D_Cg+l2*+D_WT+l3*+D_WH); A feature vector (A, M, B, S) is formed by using at least one of the action deviation A, the abnormal value M, the tension index B, and the risk factor S; Step S302: Calculate the inherent risk value R of each node based on the composed feature vector (A, M, B, S) base , characterized by: R base =αA+βM+γB+δS, where α, β, γ, and δ are weight coefficients, and normalization ensures R base ∈[0,1]; Combined with the teaching experiment risk map, the spatiotemporal attention is integrated to calculate the causal chain risk conduction; Furthermore, step S302 is specifically as follows: Step S3021: locating the corresponding node of the teaching experiment risk map according to the current operation step; Step S3022: recursively traverse all upstream operations of the node and calculate the precursor risk transmission amount; Starting from the current node, recursively operate upstream, and the formula is represented as: R cond (u)=R base (u)+Σ p∈parents(u) w p→u ×R cond (p) × T(p); Among them, R cond (u) represents the conduction risk value of node u; parents(u) represents the set of direct predecessor nodes of node u, and p represents the elements in the set of direct predecessor nodes; R base (u) represents the inherent risk of node u; w p→u represents the transmission weight from node p to node u; R cond (p) represents the conduction risk value of node p; T(p) represents the time attenuation factor of node p; The calculation formula of precursor risk transmission is represented as follows: P risk =Σ u∈parents(v) w u→v ×R cond (u)×T(u); where, P risk represents the predecessor conduction risk; parents(v) represents the set of direct predecessor nodes of node v; T(u) represents the time attenuation factor of node u; wu→v represents the transmission weight from node u to node v; When the operation interval time t is greater than the preset threshold, the -λt Calculate the time decay factor; when the operation interval time t is within the preset threshold, the time decay factor is always 1; Wherein, e represents the natural constant; t represents the operation interval time; λ represents the adjustment coefficient; It should be noted that the calculation formula for the precursor risk transmission amount aggregates the transmission risks of all direct predecessor nodes and multiplies them by time decay. The recursive process ensures that indirect predecessor risks are included, making the calculation of the precursor risk transmission amount more reasonable; If node u has no predecessor node (i.e. source node), then R cond (u)=R base (u); Step S3023: Based on the result severity matrix W, the severity of all possible results is accumulated, and the formula is: S=Σ k=1 K W v,k ; Where S represents the severity of the result, ranging from [0, K]; W v,k represents the severity value of the kth possible outcome of node v in the outcome severity matrix W, W v,k ∈[0,1]; K represents the number of possible outcomes of node v; It should be noted that the resulting severity matrix W is predefined, and each entry W v,k Represents a specific result. Since the severity matrix W is different for different experiments, the entry W can be obtained according to the number of experimental steps. v,k The number of , usually does not require a lot of calculation; In this embodiment, the specific results include "reagent leakage" and "reactor rupture", etc. The results are enumerable and the severity matrix W can be easily preset; Step S3024: Combine the precursor risk transmission amount and the current result severity to generate a transmission risk score of 0-1, specifically: Risk score =z×P risk +(1-z)×[S / S max ]; Among them, Risk score represents the conduction risk score; z represents the adjustment coefficient, which is used to coordinate the precursor risk conduction amount and the current result severity, z∈[0,1]; S max Indicates the maximum result severity of the node; Furthermore, in step S400, a GAN risk scenario simulation is performed based on the multimodal data acquired in real time; the risk gain is calculated based on the GAN risk scenario simulation results, and a graded warning is performed based on the causal chain risk transmission calculation and the risk gain; Step S401: pre-build a conditional GAN ​​model through deep learning, and input a noise vector and a real-time feature vector (A, M, B, S) into the conditional GAN ​​model generator; Inputting historical risk data into the conditional GAN ​​model discriminator; Step S402: Output the simulated feature vector (A) at the output end of the conditional GAN ​​model generator. sim ,M sim ,B sim ,S sim ); Step S403: derive the simulated feature vectors (A) from several groups of simulated outputs. sim ,M sim ,B sim ,S sim ) Quantify the risk increment of the simulated scenario compared to the current state, represented by the formula: Risk_Gain=Σ N max(0,Risk sim -Risk score ) / N; Among them, Risk_Gain represents the risk increment; N represents the number of GAN model groups; Risk sim represents the simulated conduction risk score calculated based on the simulated eigenvector; It should be noted that the risk increment is a relative increment, which only accumulates positive risks, ignores scenarios where the risk is reduced, and normalizes the results; Step S404: Trigger a graded warning based on the current conduction risk score and the simulated risk increment; In this embodiment, the first risk warning threshold is preset to 0.5. When the first risk warning threshold is exceeded, a real-time warning is given to the experimenter. The second risk warning threshold is preset to 0.7. When the second risk warning threshold is exceeded, a real-time warning is given to the responsible teacher. A dynamic teaching resource management system based on digital and intelligent interaction, including: data acquisition module, data retrieval module, calculation module, GAN simulation module and early warning module; The data acquisition module is used to acquire multimodal data in real time, wherein the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; The data retrieval module is used to preset a causal chain network of experimental operations, construct a teaching experiment risk map based on the causal chain network, store the teaching experiment risk map, and provide a retrieval function; The calculation module is responsible for calculation and realizes the automated calculation process through programs; The GAN simulation module is used to pre-build a conditional GAN ​​model and simulate GAN risk scenarios based on multimodal data acquired in real time; The early warning module is used to quantify the risk increment of the simulated scenario compared with the current state through several sets of simulated feature vectors output by simulation, and trigger a graded early warning based on the current transmission risk score and the simulated risk increment.

[0014] Specifically: the computing module supports independent computing.

[0015] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A dynamic management method for teaching resources based on digital and intelligent interaction, characterized by: The teaching resource dynamic management method comprises: Step S100: sensing multimodal data, where the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; Step S200: Preset a causal chain network of experimental operations, and construct a teaching experiment risk map based on the causal chain network; Step S300: Acquire multimodal data in real time, generate feature vectors, combine the teaching experiment risk map, and integrate spatiotemporal attention to perform causal chain risk conduction calculation; Step S400: Perform a GAN risk scenario simulation based on the multimodal data acquired in real time; calculate the risk gain based on the GAN risk scenario simulation results, and perform a graded warning based on the causal chain risk transmission calculation and risk gain.

2. A method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 1, characterized in that: In step S100, multimodal data is sensed, wherein the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data, specifically: A camera is installed on the teaching laboratory table to continuously take instantaneous image snapshots during the experiment, wherein the instantaneous image snapshots include experimental operation actions, tool positions, and the addition process of chemical raw materials; The experimental instruments are equipped with sensors, including pressure sensors, temperature sensors and speed sensors, to obtain pressure, instrument surface temperature and instrument speed during the experiment; An intelligent experimental glove equipped with a pressure sensor for obtaining the pressure distribution and trembling frequency of the experimenter's hand during the experiment; Several sensors are installed in the laboratory, including temperature sensors, humidity sensors, and gas concentration sensors, to collect laboratory environmental data.

3. The method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 2, characterized in that: In step S200, a causal chain network of experimental operations is preset, specifically: The causal chain network of experimental operations is pre-set by professionals; The specific steps of constructing the teaching experiment risk map based on the causal chain network are as follows: The causal chain is defined as a five-tuple: C = [V, E, P, W, T]; where C represents the five-tuple identifier of an experiment; V represents the node set; E represents the directed edge set; P represents the transition probability matrix; W represents the result severity matrix; and T represents the time decay factor. The nodes include operations, risks and results; Including: V={Vi|i=1,2,…,n}, i represents the number of nodes, n represents the total number of nodes; The directed edges represent causal relationships, E⊆V×V; The transition probability matrix and the result severity matrix are obtained according to the causal chain network of the preset experimental operation.

4. The method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 3, characterized in that: Specifically: According to the five-tuple: C = [V, E, P, W, T], a teaching experiment risk map is constructed, specifically: The node set V in the quintuple is used as the node of the teaching experiment risk map, and the directed edge set E in the quintuple is used as the edge of the teaching experiment risk map; each edge e u→v represents a directed edge from node u to node v, u, v∈V; each edge e u→v Associate a conduction weight w u→v , which is determined based on the ratio of all errors in the history from the operation steps corresponding to node u to the operation steps corresponding to node v, and the range is (0,1).

5. The method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 4, characterized in that: In step S300, multimodal data is acquired in real time to generate feature vectors. The causal chain risk conduction calculation is performed by integrating spatiotemporal attention with the teaching experiment risk map, specifically: Step S301: During the experiment, instantaneous image snapshots are acquired in real time, processed by a pre-trained 3D-CNN model, the operator's skeletal key points and tool / reagent positions are extracted, and the action deviation A is calculated based on the Hausdorff distance with the standard template; Through the experiment, the sensor equipped on the instrument is used to read the sensor data value of the experimental instrument in real time, and the abnormal value M is calculated based on the deviation of the preset safety range; Through the smart experimental gloves, the pressure distribution and tremor frequency of the experimenter's hand are obtained in real time, and the current experimenter's tension index B is calculated based on historical data; Through various sensors installed in the laboratory, sensor data values ​​are obtained in real time to calculate the weighted risk coefficient S of indicators such as gas concentration, temperature and humidity; A feature vector (A, M, B, S) is formed by using at least one of the action deviation A, the abnormal value M, the stress index B, and the risk factor S; Step S302: Calculate the inherent risk value R of each node based on the composed feature vector (A, M, B, S) base , characterized by: R base =αA+βM+γB+δS, where α, β, γ, and δ are weight coefficients, and normalization ensures R base ∈[0,1]; combined with the teaching experiment risk map, the spatiotemporal attention is integrated to perform causal chain risk conduction calculation.

6. The method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 5, characterized in that: include: Step S3021: locating the corresponding node of the teaching experiment risk map according to the current operation step; Step S3022: recursively traverse all upstream operations of the node and calculate the precursor risk transmission amount; Starting from the current node, recursively operate upstream, and the formula is represented as: R cond (u)=R base (u)+Σ p∈parents(u) w p→u ×R cond (p)×T(p); Among them, R cond (u) represents the conduction risk value of node u; parents(u) represents the set of direct predecessor nodes of node u, and p represents the elements in the set of direct predecessor nodes; R base (u) represents the inherent risk of node u; w p→u represents the transmission weight from node p to node u; R cond (p) represents the conduction risk value of node p; T(p) represents the time attenuation factor of node p; The calculation formula of the precursor risk transmission amount is characterized as follows: P risk =Σ u∈parents(v) w u→v ×R cond (u)×T(u); where, P risk represents the predecessor conduction risk; parents(v) represents the set of direct predecessor nodes of node v; T(u) represents the time attenuation factor of node u; w u→v represents the transmission weight from node u to node v; When the operation interval time t is greater than the preset threshold, the -λt Calculate the time decay factor; when the operation interval time t is within the preset threshold, the time decay factor is always 1; Wherein, e represents the natural constant; t represents the operation interval time; λ represents the adjustment coefficient; Step S3023: Based on the result severity matrix W, the severity of all possible results is accumulated, and the formula is: S=Σ k= 1 K W v,k ; Where S represents the severity of the result, ranging from [0, K]; W v,k represents the severity value of the kth possible outcome of node v in the outcome severity matrix W, W v,k ∈[0,1]; K represents the number of possible outcomes of node v; Step S3024: Combine the precursor risk transmission amount and the current result severity to generate a transmission risk score of 0-1, specifically: Risk score =z×P risk +(1-z)×[S / S max ]; Among them, Risk score represents the conduction risk score; z represents the adjustment coefficient, which is used to coordinate the precursor risk conduction amount and the current result severity, z∈[0,1]; S max Indicates the maximum result severity of the node.

7. The method for dynamic management of teaching resources based on digital and intelligent interaction according to claim 6, characterized in that: In step S400, a GAN risk scenario simulation is performed based on the multimodal data acquired in real time, specifically: Step S401: pre-build a conditional GAN ​​model through deep learning, and input a noise vector and a real-time feature vector (A, M, B, S) into the conditional GAN ​​model generator; Inputting historical risk data into the conditional GAN ​​model discriminator; Step S402: Output the simulated feature vector (A) at the output end of the conditional GAN ​​model generator. sim ,M sim ,B sim ,S sim ); Step S403: derive the simulated feature vectors (A) from several groups of simulated outputs. sim ,M sim ,B sim ,S sim ) Quantify the risk increment of the simulated scenario compared to the current state, represented by the formula: Risk_Gain=Σ N max(0,Risk sim -Risk score ) / N; Among them, Risk_Gain represents the risk increment; N represents the number of GAN model groups; Risk sim represents the simulated conduction risk score calculated based on the simulated eigenvector; Step S404: Trigger a graded warning based on the current conduction risk score and the simulated risk increment.

8. A dynamic teaching resource management system based on digital and intelligent interaction, applying the dynamic teaching resource management method based on digital and intelligent interaction according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, data retrieval module, calculation module, GAN simulation module and early warning module; The data acquisition module is used to acquire multimodal data in real time, wherein the multimodal data includes at least one of instantaneous image snapshots, sensor data of experimental instruments, sensor data of experimenters, and environmental data; The data retrieval module is used to preset a causal chain network of experimental operations, construct a teaching experiment risk map based on the causal chain network, store the teaching experiment risk map, and provide a retrieval function; The calculation module is responsible for calculation and realizes the automatic calculation process through the program; The GAN simulation module is used to pre-build a conditional GAN ​​model and simulate GAN risk scenarios based on multimodal data acquired in real time; The early warning module is used to quantify the risk increment of the simulated scenario compared with the current state through several groups of simulated feature vectors output by simulation, and trigger a graded early warning by combining the current conduction risk score and the simulated risk increment.

9. A dynamic teaching resource management system based on digital and intelligent interaction according to claim 8, characterized in that: Specifically: the computing module supports independent computing.

Images