Chemical laboratory risk early warning method and system based on multi-modal data fusion algorithm

Abstract

The present application relates to the technical field of intelligent safety monitoring, in particular to a chemical laboratory risk early warning method and system based on a multi-modal data fusion algorithm, in the present application, the spatial gradients of multi-modal data such as gas concentration and temperature are calculated, and the coupling relationship and time delay among these gradients are analyzed, thereby realizing deep insight into the dynamic changes of the laboratory environment, no longer relying on isolated single-point threshold judgment, but evaluating stability according to the cooperative phase of heat diffusion and wet flow process, this way can effectively identify the potential unstable state caused by the combined action of multiple factors, even if each single parameter has not reached the alarm limit value, the risk can also be captured in advance, at the same time, through the aggregation analysis of adjacent unstable areas in space, the spreading trend and influence range of the risk can be revealed, thereby generating more accurate and forward-looking early warning, significantly reducing false alarms caused by non-dangerous transient fluctuations.

Description

Technical Field

[0001] This invention relates to the field of intelligent safety monitoring technology, and in particular to a method and system for early warning of chemical laboratory risks based on a multimodal data fusion algorithm. Background Technology

[0002] The field of intelligent safety monitoring technology involves using advanced technologies to monitor and warn of potential risks in various environments in real time, ensuring the safety of personnel and equipment. This field encompasses multiple technological directions, including but not limited to applications such as video surveillance, sensor networks, data analytics, artificial intelligence, and machine learning, to achieve real-time detection and early warning of safety hazards.

[0003] Among these, the chemical laboratory risk early warning method refers to using traditional monitoring technologies to predict and warn of potential risks in chemical laboratories. This method typically involves setting up fixed monitoring equipment, such as temperature, humidity, and gas concentration sensors, to collect environmental parameters in the laboratory in real time. Based on the collected data, the system will issue alarms to remind laboratory staff to take timely measures.

[0004] Existing risk warning methods in chemical laboratories mainly rely on independent monitoring of single environmental parameters. Their operation mode is limited to directly comparing the collected discrete data points with preset static thresholds. This mechanism ignores the inherent dynamic correlation between different physical quantities and cannot deeply analyze, for example, the interaction between heat changes and gas diffusion in space. Therefore, it performs poorly in distinguishing between instantaneous fluctuations caused by normal operations and the continuous evolution of real potential risks. This lack of judgment ability not only easily triggers false alarms due to local harmless disturbances, reducing the credibility of warnings, but more importantly, for some slowly developing but highly dangerous abnormal conditions, since no single indicator touches the alarm line, the risks cannot be identified in advance, ultimately weakening the foresight and accuracy of warnings. Summary of the Invention

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a chemical laboratory risk early warning method based on a multimodal data fusion algorithm, comprising the following steps:

[0006] S1: Acquire gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculate the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establish a multimodal synchronous monitoring dataset.

[0007] S2: Based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, calculate the thermal diffusion coupling parameters and the wet flow coupling parameters, determine the target spatial location with obvious coupling, and extract the local coupling change data at the target spatial location;

[0008] S3: Analyze the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupling change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and perform fusion matching to obtain temporal phase coupling data;

[0009] S4: Determine the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on the time-series phase coupling data, and integrate them into a multimodal risk distribution dataset;

[0010] S5: Perform regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk warning results.

[0011] As a further aspect of the present invention, the multimodal synchronous monitoring dataset includes concentration gradient distribution, temperature gradient distribution, humidity gradient distribution, and airflow velocity gradient distribution. The local coupled change data specifically refers to the concentration change rate, temperature change rate, humidity change rate, and airflow velocity change rate at the target spatial location. The temporal phase coupled data includes thermal diffusion time delay, moisture flow time delay, spatial coordinates, and time markers. The multimodal risk distribution dataset includes information on cooperatively stable phase regions, information on potentially unstable phase regions, and phase difference parameters corresponding to the regions. The risk warning result specifically refers to the risk spatial location and risk level.

[0012] As a further aspect of the present invention, step S1 specifically comprises:

[0013] S101: Acquire raw monitoring data collected by gas concentration sensors, temperature sensors, humidity sensors and airflow velocity sensors at designated locations in the chemical laboratory, and perform synchronization alignment based on sensor timestamps. Record the gas concentration, temperature, humidity and airflow velocity at the same time point, associate them with the three-dimensional spatial coordinates marked on the sensor layout map, and generate a monitoring data spatial coordinate mapping set.

[0014] S102: Based on the monitoring data spatial coordinate mapping set, determine the adjacent monitoring points in the three-dimensional spatial coordinate system, call the coordinate data of the adjacent monitoring points to calculate the Euclidean spatial distance between the two points, and obtain the gas concentration difference, temperature difference, humidity difference and airflow velocity difference of the adjacent monitoring points at the same time stamp. Divide each difference by the corresponding spatial distance to obtain the multidimensional monitoring data spatial gradient value.

[0015] S103: Call the monitoring data spatial coordinate mapping set and the multi-dimensional monitoring data spatial gradient value, and establish a joint index for gas concentration, temperature, humidity, airflow velocity and corresponding gradient data in a unified coordinate system to obtain a multi-modal synchronous monitoring dataset.

[0016] As a further aspect of the present invention, step S2 specifically includes:

[0017] S201: Extract the direction parameters of the concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient in the multimodal synchronous monitoring data set, calculate the angle between the directions of the temperature gradient and the concentration gradient for the same spatial coordinate position to obtain the thermal diffusion coupling parameter, and calculate the angle between the directions of the humidity gradient and the airflow velocity gradient to obtain the moisture flow coupling parameter, and establish a thermal-humidity coupling parameter set.

[0018] S202: Call the preset thermal diffusion coupling threshold and wet flow coupling threshold, and based on the comparison of the thermal and wet flow coupling parameter set, filter the spatial locations where the thermal diffusion coupling parameter is less than the thermal diffusion coupling threshold and the wet flow coupling parameter is less than the wet flow coupling threshold, as the target spatial locations with obvious coupling;

[0019] S203: Extract the concentration change rate, temperature change rate, humidity change rate and airflow velocity change rate corresponding to the target spatial location with obvious coupling from the multimodal synchronous monitoring dataset, and establish local coupling change data.

[0020] As a further aspect of the present invention, step S3 specifically comprises:

[0021] S301: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the gas concentration change rate sequence and the temperature change rate sequence at the target spatial location, and obtain the thermal diffusion time delay;

[0022] S302: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the humidity change rate sequence and the airflow velocity change rate sequence of the target spatial location, and obtain the wet flow time delay;

[0023] S303: Invoke the time sequence and spatial coordinates corresponding to the heat diffusion time delay and the wet flow time delay to establish time-phase coupled data.

[0024] As a further aspect of the present invention, step S4 specifically comprises:

[0025] S401: Compare the thermal diffusion time delay and the wet flow time delay in the time-series phase coupling data with their respective thermal diffusion thresholds and preset wet flow thresholds to generate phase region calibration results;

[0026] S402: Based on the spatial locations calibrated in the phase region calibration results, find and extract the concentration gradient data, temperature gradient data, humidity gradient data, and airflow velocity gradient data corresponding to these locations from the multimodal synchronous monitoring dataset, and establish a target region gradient dataset;

[0027] S403: Based on the common spatial location information in the phase region calibration result and the target region gradient dataset, a multimodal risk distribution dataset is established.

[0028] As a further aspect of the present invention, the process of comparing the respective thermal diffusion threshold with the preset moisture flow threshold includes:

[0029] Iterate through each data record in the time-series phase-coupled data and extract the thermal diffusion time delay and wet flow time delay from the data record;

[0030] The extracted heat diffusion time delay is compared with the heat diffusion threshold, and the extracted wet flow time delay is compared with the preset wet flow threshold.

[0031] If the thermal diffusion time delay is less than the thermal diffusion threshold and the wet flow time delay is less than the preset wet flow threshold, then the spatial location associated with the data record is marked as the cooperative stable phase region.

[0032] If the thermal diffusion time delay is greater than or equal to the thermal diffusion threshold, or the wet flow time delay is greater than or equal to the preset wet flow threshold, then the spatial location associated with the data record will be marked as a potentially unstable phase region.

[0033] As a further aspect of the present invention, step S5 specifically includes:

[0034] S501: Extract potential unstable phase regions within the same time window from the multimodal risk distribution dataset, calculate the spatial proximity relationship between each potential unstable phase region, determine the degree of aggregation association between adjacent regions based on coordinate distance and gradient magnitude weight, and establish a set of spatially aggregated and associated regions.

[0035] S502: Based on the multimodal risk distribution dataset, obtain the thermal diffusion time delay and gradient data within the spatially aggregated associated region set, calculate the average thermal diffusion phase difference and average wet flow phase difference of each aggregated region in the spatially aggregated associated region set, determine the consistency between the phase difference change trend and the gradient change direction within the aggregated region in the spatially aggregated associated region set, and generate the aggregated region risk assessment result.

[0036] S503: Compare the average thermal diffusion phase difference and the average wet flow phase difference of the risk assessment results of the aggregation area with their respective risk judgment thresholds. When any average phase difference exceeds the corresponding risk judgment threshold, generate a record on the corresponding spatial coordinates of the aggregation area and establish a risk warning result.

[0037] As a further aspect of the present invention, the process of determining the consistency between the phase difference change trend and the gradient change direction within the spatially aggregated region includes:

[0038] Traverse each aggregation region in the spatial aggregation association region set and extract continuous time series data of thermal diffusion time delay, moisture flow time delay, concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient within the aggregation region;

[0039] Based on the continuous time series, calculate the trend vectors of the thermal diffusion time delay and the wet flow time delay;

[0040] Based on the continuous time series, calculate the combined change direction vectors of temperature gradient and concentration gradient, and the combined change direction vectors of humidity gradient and airflow velocity gradient, respectively.

[0041] Calculate the angle between the trend vector of the thermal diffusion time delay and the combined change direction vector of the temperature gradient and concentration gradient, and compare the angle with a preset consistency angle threshold.

[0042] Simultaneously, the angle between the trend vector of the change in wet flow time delay and the combined change direction vector of humidity gradient and airflow velocity gradient is calculated, and the angle is compared with the consistency angle threshold.

[0043] When the angle calculated by both types is less than the consistency angle threshold, the phase difference change trend is determined to be consistent with the gradient change direction.

[0044] A chemical laboratory risk early warning system based on multimodal data fusion algorithms includes:

[0045] The data gradient module acquires gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculates the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establishes a multimodal synchronous monitoring dataset.

[0046] The spatial coupling positioning module calculates thermal diffusion coupling parameters and moisture flow coupling parameters based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, determines the target spatial location with obvious coupling, and extracts local coupling change data at the target spatial location.

[0047] The phase coupling analysis module analyzes the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupling change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and performs fusion matching to obtain time-series phase coupling data.

[0048] The risk area identification module determines the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on the time-series phase coupling data, and integrates them into a multimodal risk distribution dataset.

[0049] The risk aggregation and early warning module performs regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk early warning results.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0051] This invention achieves a deep understanding of the dynamic changes in the laboratory environment by calculating the spatial gradients of multimodal data such as gas concentration and temperature, and analyzing the coupling relationship and time delay between these gradients. It no longer relies on isolated single-point threshold judgments, but assesses stability based on the synergistic phase of thermal diffusion and moisture flow processes. This approach can effectively identify potential unstable states caused by the combined effects of multiple factors. Even if each individual parameter has not yet reached the alarm limit, it can capture the nascent risks in advance. At the same time, by performing aggregate analysis on spatially adjacent unstable regions, it can reveal the spread trend and scope of impact of risks, thereby generating more accurate and forward-looking early warnings, significantly reducing false alarms caused by non-hazardous instantaneous fluctuations, and improving the ability to detect complex and progressive risks early. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of the steps of the present invention;

[0054] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0055] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0056] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0057] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0058] Figure 6 This is a detailed schematic diagram of S5 of the present invention;

[0059] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0060] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0061] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0062] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0063] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0064] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0065] Please see Figure 1 This invention provides a method for risk warning in chemical laboratories based on a multimodal data fusion algorithm, comprising the following steps:

[0066] S1: Acquire gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculate the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establish a multimodal synchronous monitoring dataset.

[0067] S2: Based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, calculate the thermal diffusion coupling parameters and the wet flow coupling parameters, determine the target spatial location with obvious coupling, and extract the local coupling change data at the target spatial location;

[0068] S3: Analyze the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupled change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and perform fusion matching to obtain temporal phase coupled data;

[0069] S4: Determine the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on time-series phase coupling data, and integrate them into a multimodal risk distribution dataset;

[0070] S5: Perform regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk warning results;

[0071] The multimodal synchronous monitoring dataset includes concentration gradient distribution, temperature gradient distribution, humidity gradient distribution, and airflow velocity gradient distribution. The local coupled change data specifically includes the rate of change of concentration, temperature, humidity, and airflow velocity at the target spatial location. The temporal phase coupled data includes thermal diffusion time delay, moisture flow time delay, spatial coordinates, and time stamps. The multimodal risk distribution dataset includes information on cooperatively stable phase regions, information on potentially unstable phase regions, and the corresponding phase difference parameters for each region. The risk warning results specifically refer to the spatial location and risk level of the risk.

[0072] Please see Figure 2 The specific steps of S1 are as follows:

[0073] S101: Acquire raw monitoring data collected by gas concentration sensors, temperature sensors, humidity sensors and airflow velocity sensors at designated locations in the chemical laboratory, and perform synchronization alignment based on sensor timestamps. Record the gas concentration, temperature, humidity and airflow velocity at the same time point, associate them with the three-dimensional spatial coordinates marked on the sensor layout map, and generate a monitoring data spatial coordinate mapping set.

[0074] Specific locations were selected within the chemistry laboratory, including near the fume hood, the chemical cabinet area, and the central workbench, to deploy monitoring equipment. Specifically, monitoring terminals integrating ammonia concentration sensors, digital temperature and humidity sensors, and thermal airflow velocity sensors were deployed at the following locations: 1 meter directly in front of the fume hood at a height of 1.5 meters (coordinates (1, 4, 1.5)); 0.5 meters in front of the chemical cabinet at a height of 1.2 meters (coordinates (5, 1, 1.2)); and the center of the central workbench at a height of 1.5 meters (coordinates (3, 3, 1.5)). Each monitoring terminal continuously collected data at a frequency of 1 Hz, and each set of collected data was appended with a UNIX timestamp accurate to milliseconds. During a certain collection cycle, the data packet obtained from monitoring point one, with timestamp 1677628800.105, contained the following information: ammonia concentration 5.2 ppm, temperature 21.5 degrees Celsius, relative humidity 45.3%, and airflow velocity 0.15 m / s. The data packet timestamp from monitoring point two is 1677628800.108, recording an ammonia concentration of 15.8 ppm, a temperature of 21.3 degrees Celsius, a relative humidity of 46.1%, and an airflow velocity of 0.05 m / s. The data packet timestamp from monitoring point three is 1677628800.106, recording an ammonia concentration of 7.5 ppm, a temperature of 21.6 degrees Celsius, a relative humidity of 45.5%, and an airflow velocity of 0.12 m / s. A synchronization alignment window of 5 milliseconds is set, and data with timestamp differences within this window are considered to be collected at the same time point. The maximum difference in timestamps among the three data packets is 3 milliseconds, which is less than the 5-millisecond window, and therefore they are grouped into the same synchronization period. Subsequently, the multidimensional monitoring data after synchronization alignment is correlated with the pre-calibrated sensor three-dimensional spatial coordinates. This correlation operation generates a structured data record for each synchronization time point, containing the timestamp of that time point and a set consisting of the coordinates of each monitoring point and its corresponding multidimensional monitoring data. For example, using 1677628800.106 as the base timestamp, the generated records are: {timestamp: 1677628800.106, data points: [{coordinates: (1, 4, 1.5), ammonia concentration: 5.2 ppm, temperature: 21.5℃, humidity: 45.3%RH, airflow velocity: 0.15 m / s}, {coordinates: (5, 1, 1.2), ammonia concentration: 15.8 ppm, temperature: 21.3℃, humidity: 46.1%RH, airflow velocity: 0.05 m / s}, {coordinates: (3, 3, 1.5), ammonia concentration: 7.5 ppm, temperature: 21.6℃, humidity: 45.5%RH, airflow velocity: 0.12 m / s}]}. By performing this synchronization and correlation operation on the continuous data stream, a time-series-based monitoring data spatial coordinate mapping set is finally generated.

[0075] S102: Based on the spatial coordinate mapping set of monitoring data, determine the adjacent monitoring points in the three-dimensional spatial coordinate system, call the coordinate data of the adjacent monitoring points to calculate the Euclidean spatial distance between the two points, and obtain the gas concentration difference, temperature difference, humidity difference and airflow velocity difference of the adjacent monitoring points at the same time stamp. Divide each difference by the corresponding spatial distance to obtain the multidimensional monitoring data spatial gradient value.

[0076] The spatial coordinate mapping set of monitoring data is retrieved to determine adjacent monitoring points in a three-dimensional spatial coordinate system. Adjacency is determined by a preset spatial distance threshold, which is established by analyzing historical environmental simulation data from similar chemical laboratories to identify the maximum distance between sensor readings that still maintains a significant statistical correlation; in this embodiment, it is set to 4 meters. For example, the Euclidean spatial distance between monitoring point one (coordinates 1, 4, 1.5) and monitoring point three (coordinates 3, 3, 1.5) needs to be calculated. The calculation process first obtains the square of the difference between the two points on the horizontal axis, i.e., (3-1) squared, which results in 4; then, the square of the difference between the two points on the vertical axis, i.e., (3-4) squared, which results in 1; and then the square of the difference between the two points on the vertical axis, i.e., (1.5-1.5) squared, which results in 0. These three squared values ​​are added together, resulting in 4 + 1 + 0 = 5. Finally, the square root of 5 is calculated, yielding approximately 2.24 meters. This distance is less than the 4-meter threshold; therefore, monitoring point one and monitoring point three are determined to be adjacent monitoring points. This distance calculation and comparison process is repeated for all monitoring points. Next, the differences in monitoring data at the same timestamp are obtained for the monitoring points identified as adjacent. Taking the data of monitoring point 1 and monitoring point 3 at timestamp 1677628800.106 as an example, the gas concentration difference is 7.5 ppm at monitoring point 3 minus 5.2 ppm at monitoring point 1, resulting in 2.3 ppm. The temperature difference is 21.6 degrees Celsius at monitoring point 3 minus 21.5 degrees Celsius at monitoring point 1, resulting in 0.1 degrees Celsius. The humidity difference is 45.5%RH at monitoring point 3 minus 45.3%RH at monitoring point 1, resulting in 0.2%RH. The airflow velocity difference is 0.12 m / s at monitoring point 3 minus 0.15 m / s at monitoring point 1, resulting in -0.03 m / s. Finally, the difference in each monitoring data is divided by the corresponding spatial distance to obtain the spatial gradient value of the multidimensional monitoring data. The spatial gradient of gas concentration is 2.3 ppm divided by 2.24 meters, approximately 1.027 ppm / m. The spatial gradient of temperature is 0.1 degrees Celsius divided by 2.24 meters, approximately 0.045 degrees Celsius / m. The spatial gradient of humidity is 0.2%RH divided by 2.24 meters, approximately 0.089%RH / m. The spatial gradient of airflow velocity is -0.03 m / s divided by 2.24 meters, approximately -0.013 (m / s) / m.

[0077] S103: Call the monitoring data spatial coordinate mapping set and the multi-dimensional monitoring data spatial gradient value, and establish a joint index for gas concentration, temperature, humidity, airflow velocity and corresponding gradient data under a unified coordinate system to obtain a multi-modal synchronous monitoring dataset;

[0078] This process involves calling upon the spatial coordinate mapping set of monitoring data and the spatial gradient values ​​of multidimensional monitoring data to establish a joint index for gas concentration, temperature, humidity, airflow velocity, and their corresponding gradient data within a unified coordinate system. The aim is to structurally integrate raw state monitoring data with gradient data characterizing their spatial rate of change. Specifically, it involves expanding each timestamp record in the spatial coordinate mapping set of monitoring data. The original record contains the coordinates of each monitoring point and its instantaneous four-dimensional monitoring data. The expansion operation adds an association field to each pair of monitoring points determined to be adjacent, storing the calculated four-dimensional spatial gradient value between them. For example, when processing data with timestamp 1677628800.106, the process identifies monitoring point 1 and monitoring point 3 as adjacent. Therefore, in the data structure for this timestamp, in addition to recording the original monitoring data for each of monitoring points 1 and 3, an additional gradient data entry associated with the pair of monitoring points 1 and 3 is generated. This entry contains: {Associated point pairs: [Monitoring point 1 (coordinates 1, 4, 1.5), Monitoring point 3 (coordinates 3, 3, 1.5)], Concentration gradient: 1.027 ppm / m, Temperature gradient: 0.045℃ / m, Humidity gradient: 0.089%RH / m, Airflow velocity gradient: -0.013 (m / s) / m}. The composite index is built using a hash table structure. Its key consists of a timestamp and spatial coordinates (or coordinate pairs), while its value is the corresponding original monitoring dataset or gradient dataset. When querying the state of a specific location at a specific time, the timestamp and the coordinates of that location are used as the index to directly retrieve the original data such as gas concentration and temperature at that point. When querying the degree of environmental change near that location, the timestamp and the coordinates of adjacent point pairs containing that location are used as the index to retrieve the corresponding gradient data. In this way, the "point" data describing the state and the "edge" data describing the trend of change are strongly correlated at the database level, ultimately forming a structured multimodal synchronous monitoring dataset.

[0079] Please see Figure 3 The specific steps of S2 are as follows:

[0080] S201: Extract the directional parameters of concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient in the multimodal synchronous monitoring data. Calculate the angle between the directions of the temperature gradient and concentration gradient for the same spatial coordinate position to obtain the thermal diffusion coupling parameter, and calculate the angle between the directions of the humidity gradient and airflow velocity gradient to obtain the moisture flow coupling parameter. Establish a thermal-humidity coupling parameter set.

[0081] From the multimodal synchronous monitoring dataset, the direction parameters of the concentration gradient, temperature gradient, humidity gradient, and airflow velocity gradient around a specific spatial coordinate location are extracted; that is, the direction cosines of the gradient vectors in three-dimensional space. Taking monitoring point three (coordinates 3, 3, 1.5) as an example, assume that the comprehensive temperature gradient vector calculated from multiple adjacent points around it is (0.04, -0.01, 0), and the concentration gradient vector is (0.9, -0.2, 0). Then, the angle between these two vectors is calculated as the thermal diffusion coupling parameter. The calculation process is as follows: First, the dot product of the two vectors is calculated, i.e., (0.04 × 0.9) + ((-0.01) × (-0.2)) + (0 × 0), resulting in 0.038. Next, the magnitude of each vector is calculated. The magnitude of the temperature gradient vector is the square root of (0.04² + (-0.01)² + 0²), approximately equal to 0.0412. The magnitude of the concentration gradient vector is the square root of (0.9² + (-0.2)² + 0²), approximately 0.922. Then, dividing the dot product by the product of the two magnitudes, i.e., 0.038 / (0.0412 × 0.922), yields approximately 0.999. Finally, calculating the arccosine of this value gives an angle of approximately 2.5 degrees. This 2.5 degrees is the thermal diffusion coupling parameter at monitoring point three. Similarly, assuming the humidity gradient vector at monitoring point three is (0.08, 0.03, 0) and the airflow velocity gradient vector is (-0.01, 0.05, 0), the humidity-flow coupling parameter is calculated. The dot product of the two vectors is (0.08 × (-0.01)) + (0.03 × 0.05) + (0 × 0), resulting in 0.0007. The magnitude of the humidity gradient vector is approximately 0.0854, and the magnitude of the airflow velocity gradient vector is approximately 0.051. The dot product divided by the product of the moduli is 0.0007 / (0.0854×0.051), which is approximately 0.160. Its inverse cosine value is approximately 80.8 degrees. This 80.8 degrees is the moisture coupling parameter at monitoring point three. Repeating this calculation process for all monitoring points establishes a set of thermo-humidity coupling parameters covering the entire monitoring area.

[0082] S202: Call the preset thermal diffusion coupling threshold and wet flow coupling threshold, and based on the comparison of the thermal and wet flow coupling parameter set, filter the spatial locations where the thermal diffusion coupling parameter is less than the thermal diffusion coupling threshold and the wet flow coupling parameter is less than the wet flow coupling threshold, as the target spatial locations with obvious coupling.

[0083] The determination of the thermal diffusion coupling threshold is based on statistical analysis of controlled experimental data. In the experiment, benign heat sources and pollutant sources generated by various normal laboratory operations (such as switching equipment on and off, personnel movement) were simulated, and the angles between the resulting temperature gradient and concentration gradient directions were recorded. This dataset of angles was statistically analyzed, and the sum of the mean and twice the standard deviation was used as the threshold; in this example, 15 degrees was calculated. The humidity-flow coupling threshold was set using the same method. Experiments simulated normal ventilation and the start-up and shutdown of humidification equipment, and the angles between the humidity gradient and airflow velocity gradient were collected. Similarly, the sum of the mean and twice the standard deviation of this dataset was used as the threshold; in this example, 20 degrees was calculated. Experimental verification of these thresholds was conducted in a controlled environment chamber. In the experiment verifying the thermal diffusion threshold, a controllable power heating plate and an ammonia release source were arranged in the chamber. By analyzing the cumulative distribution of gradient angles under different operating conditions, it was confirmed that the value at the 95th percentile matched the calculated 15-degree threshold. In the experiment verifying the moisture-flow coupling threshold, a humidifier and a fan were placed inside the chamber. By analyzing the cumulative distribution of the gradient angle under different combinations of wind speed and humidification, it was confirmed that the value at the 95th percentile matched the calculated 20-degree threshold. The screening process was based on the thermo-humidity coupling parameter set, comparing the parameters at each spatial location one by one. For example, for monitoring point three, its thermo-diffusion coupling parameter was 2.5 degrees, and its moisture-flow coupling parameter was 80.8 degrees. Comparing 2.5 degrees with the thermo-diffusion coupling threshold of 15 degrees, 2.5 degrees is less than 15 degrees. Simultaneously, comparing 80.8 degrees with the moisture-flow coupling threshold of 20 degrees, 80.8 degrees is greater than 20 degrees. Since the moisture-flow coupling parameter did not meet the condition of being less than the threshold, monitoring point three was not selected as a target spatial location with significant coupling. If there is another location, namely monitoring point four, with a thermal diffusion coupling parameter of 10 degrees and a moisture flow coupling parameter of 18 degrees, then monitoring point four is selected as the target spatial location with obvious coupling because 10 degrees is less than 15 degrees and 18 degrees is less than 20 degrees.

[0084] S203: Extract the concentration change rate, temperature change rate, humidity change rate and airflow velocity change rate corresponding to the target spatial location with obvious coupling from the multimodal synchronous monitoring dataset, and establish local coupled change data;

[0085] From the multimodal synchronous monitoring dataset, monitoring data for selected target spatial locations with significant coupling were extracted over continuous time series. Taking monitoring point four as an example, ammonia concentration, temperature, humidity, and airflow velocity data recorded once per second over the past minute were retrieved. For instance, at a starting time point, the ammonia concentration at monitoring point four was 8.0 ppm; one second later, the concentration was 8.2 ppm; and one second after that, the concentration was 8.5 ppm. The rate of concentration change was calculated by subtracting the value from the value at the previous time point from the value at the next time point. At the second time point, the rate of concentration change was 8.2 ppm - 8.0 ppm = 0.2 ppm / s. At the third time point, the rate of concentration change was 8.5 ppm - 8.2 ppm = 0.3 ppm / s. A concentration rate of change sequence was obtained by performing point-by-point differencing on the entire time series data. Similarly, the same differencing operation was performed on the continuous time series data for the three parameters: temperature, humidity, and airflow velocity. For example, if the temperature at monitoring point four is 22.0℃, 22.1℃, and 22.3℃ at the initial, second, and third time points, respectively, then the first two values ​​of the corresponding temperature change rate sequence are 0.1℃ / s and 0.2℃ / s. The humidity at the same three time points is 50.0%RH, 50.1%, and 50.1%RH, resulting in a humidity change rate sequence of 0.1%RH / s and 0.0%RH / s. The airflow velocity at the same three time points is 0.20m / s, 0.22m / s, and 0.23m / s, resulting in an airflow velocity change rate sequence of 0.02m / s² and 0.01m / s². By integrating these four change rate data corresponding to all clearly coupled target spatial locations, a locally coupled change data set is established.

[0086] Please see Figure 4 The specific steps of S3 are as follows:

[0087] S301: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the gas concentration change rate sequence and the temperature change rate sequence at the target spatial location, and obtain the thermal diffusion time delay;

[0088] The data retrieves the gas concentration change rate sequence and temperature change rate sequence from monitoring point four, a location with significant coupling in the localized coupled change data. Assume that within a given time period, the temperature change rate sequence at monitoring point four reaches a peak of 0.5 degrees Celsius / second at the 5th second, while the gas concentration change rate sequence reaches a peak of 0.6 ppm / second at the 8th second. The time difference between these two peaks, i.e., 8 seconds minus 5 seconds, equals 3 seconds; this time difference is recorded as the thermal diffusion time delay. This calculation is achieved through cross-correlation analysis of the two change rate time series. The specific process of cross-correlation analysis involves using one time series (e.g., the temperature change rate sequence) as a reference, shifting the other time series (concentration change rate sequence) along the time axis, and calculating the correlation coefficient between the two sequences for each shift. Ultimately, the time shift that results in the maximum correlation coefficient is determined as the time delay between the two sequences. For example, when the concentration change rate sequence lags by 3 seconds in time, its waveform is most similar to that of the temperature change rate sequence, showing the highest correlation; therefore, the time delay is determined to be 3 seconds. This calculation is repeated for multiple sets of significant change events occurring at monitoring point four at different time periods, yielding a series of thermal diffusion time delay data.

[0089] S302: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the humidity change rate sequence and the airflow velocity change rate sequence of the target spatial location, and obtain the wet flow time delay;

[0090] The humidity change rate sequence and airflow velocity change rate sequence at monitoring point four, a location with significant coupling in the locally coupled change data, are used. Assuming that within the same observation period, the airflow velocity change rate sequence at monitoring point four peaks at 10 seconds (0.05 m / s²), while the humidity change rate sequence peaks at 15 seconds (0.4% RH / s), the time delay between these two peaks, 15 seconds minus 10 seconds, equals 5 seconds. This time difference is recorded as the wet flow time delay. Similar to the calculation of the heat diffusion time delay, cross-correlation analysis is used here. Using the airflow velocity change rate sequence as a baseline, the cross-correlation function of the humidity change rate sequence under different time shifts is calculated to find the time shift that maximizes the function value. For example, calculations show that when the overall humidity change rate sequence is delayed by 5 seconds, its trend best matches the airflow velocity change rate sequence, reaching a peak correlation. Therefore, the wet flow time delay is determined to be 5 seconds. Multiple sets of moisture and airflow disturbance events occurring at the target spatial location monitoring point four within different time periods were analyzed to obtain a series of moisture flow time delay values.

[0091] S303: Call the time sequence and spatial coordinates corresponding to the thermal diffusion time delay and the wet flow time delay to establish temporal phase coupling data;

[0092] The calculated thermal diffusion time delay and the acquired moisture flow time delay are fused and matched according to their temporal sequence and spatial coordinates. This process creates a comprehensive record for each analyzed event. For example, for a disturbance event occurring at monitoring point four, a target spatial location with significant coupling, the analyzed thermal diffusion time delay is 3 seconds, and the moisture flow time delay is 5 seconds. A data record is then generated with the following content: {Spatial coordinates: spatial coordinates of monitoring point four, event occurrence time: a specific time point, thermal diffusion time delay: 3 seconds, moisture flow time delay: 5 seconds}. The event occurrence time here can be defined as the time when the first physical quantity (such as the rate of temperature change) to peak in the disturbance event reaches its peak. By analyzing multiple coupled events occurring at all target spatial locations during continuous monitoring, the two time delay data calculated for each event are bound to the spatiotemporal information of the event, continuously generating such data records. These records are aggregated to form temporal phase coupling data.

[0093] Please see Figure 5 The specific steps of S4 are as follows:

[0094] S401: Compare the thermal diffusion time delay and wet flow time delay in the time-series phase coupling data with their respective thermal diffusion thresholds and preset wet flow thresholds to generate phase region calibration results;

[0095] The process of comparing each of the respective thermal diffusion thresholds with the preset moisture flow thresholds includes:

[0096] Iterate through each data record in the time-series phase-coupled data and extract the thermal diffusion time delay and wet flow time delay from the data record;

[0097] The extracted heat diffusion time delay is compared with the heat diffusion threshold, and the extracted wet flow time delay is compared with the preset wet flow threshold.

[0098] If the thermal diffusion time delay is less than the thermal diffusion threshold and the wet flow time delay is less than the preset wet flow threshold, then the spatial location associated with the data record is marked as the cooperative stable phase region.

[0099] If the thermal diffusion time delay is greater than or equal to the thermal diffusion threshold, or the wet flow time delay is greater than or equal to the preset wet flow threshold, then the spatial location associated with the data record will be marked as a potentially unstable phase region.

[0100] The thermal diffusion time delay in the time-phase coupled data is compared with the thermal diffusion threshold, and the moisture flow time delay is compared with the preset moisture flow threshold. The thermal diffusion threshold is set based on: simulating various benign operations (such as opening and closing the incubator door, normal personnel movement) in a controlled experiment, and recording the resulting thermal diffusion time delay. Statistical analysis is performed on the collected delay dataset, and the mean plus twice the standard deviation is taken as a critical value, considered the maximum acceptable delay under normal disturbance; in this embodiment, it is set to 3.5 seconds. The moisture flow threshold is set based on: simulating normal ventilation and humidification equipment start-up and shutdown operations in a controlled experiment, and recording the moisture flow time delay. Using the same statistical method (mean plus twice the standard deviation), the moisture flow threshold is set to 6.0 seconds. Each record in the time-phase coupled data is iterated through. For example, processing a record: {Spatial coordinates: spatial coordinates of monitoring point four, thermal diffusion time delay: 3 seconds, moisture flow time delay: 5 seconds}. First, the recorded heat diffusion time delay of 3 seconds is compared with the heat diffusion threshold of 3.5 seconds; 3 seconds is less than 3.5 seconds. Then, the recorded moisture flow time delay of 5 seconds is compared with the moisture flow threshold of 6.0 seconds; 5 seconds is less than 6.0 seconds. Since both conditions are met, the spatial location monitoring point four associated with this record is designated as a cooperatively stable phase region at this time point. Next, another record is processed: {spatial coordinates: coordinates of another spatial location monitoring point five, heat diffusion time delay: 4 seconds, moisture flow time delay: 5.5 seconds}. The heat diffusion time delay of 4 seconds is compared with the threshold of 3.5 seconds; 4 seconds is greater than or equal to 3.5 seconds. Since the heat diffusion time delay condition is not met, the spatial location monitoring point five associated with this record is designated as a potentially unstable phase region.

[0101] S402: Based on the spatial locations calibrated in the phase region calibration results, find and extract the concentration gradient data, temperature gradient data, humidity gradient data, and airflow velocity gradient data corresponding to these locations from the multimodal synchronous monitoring dataset to establish a target region gradient dataset.

[0102] Based on the spatial locations of all areas identified as cooperatively stable or potentially unstable phase regions in the phase region calibration results, gradient data associated with these spatial locations and corresponding time points are searched and extracted from the multimodal synchronous monitoring dataset. For example, the phase region calibration results show that at the time of an event, monitoring point four is identified as a cooperatively stable phase region, while monitoring point five is identified as a potentially unstable phase region. Based on this result, two queries are initiated to the multimodal synchronous monitoring dataset. The parameters of the first query are the time and spatial location monitoring point four, and the dataset returns the concentration gradient, temperature gradient, humidity gradient, and airflow velocity gradient data of the adjacent point pairs of monitoring point four at the time of the event. Assume the returned gradient data is: {Associated point pair: [monitoring point four, and another adjacent monitoring point], concentration gradient: 0.8 ppm / m, temperature gradient: 0.1℃ / m, humidity gradient: 0.2%RH / m, airflow velocity gradient: 0.03 (m / s) / m}. The second query parameters are time and spatial location monitoring point five, returning four gradient data points for monitoring point five at the time of the event. Assume the returned data is: {Associated point pair: [monitoring point five, and another adjacent monitoring point], concentration gradient: 3.5 ppm / m, temperature gradient: 0.5℃ / m, humidity gradient: 0.1%RH / m, airflow velocity gradient: -0.05(m / s) / m}. All these gradient data points extracted based on the calibration results are combined to form the target area gradient dataset.

[0103] S403: Establish a multimodal risk distribution dataset by associating the common spatial location information in the phase region calibration results and the target region gradient dataset;

[0104] The process correlates the spatial location information shared by the phase region calibration results and the target region gradient dataset. This directly links the phase state (stable or unstable) of a location with the degree of environmental change (gradient data) at that location. Specifically, it iterates through each record in the phase region calibration results. For each record, such as {spatial location: monitoring point five, calibration result: potentially unstable phase region}, its spatial location information "monitoring point five" is used as an index to search for all gradient data records related to monitoring point five in the target region gradient dataset. The found records are {associated point pair: [monitoring point five, another adjacent monitoring point], concentration gradient: 3.5 ppm / m, temperature gradient: 0.5℃ / m, ...}. Then, these two pieces of information are merged into a new, more complete data record. The new recording structure is as follows: {Spatial coordinates: Monitoring point 5, Phase state: Potentially unstable, Concentration gradient: 3.5 ppm / m, Temperature gradient: 0.5℃ / m, Humidity gradient: 0.1%RH / m, Airflow velocity gradient: -0.05 (m / s) / m, Thermal diffusion time delay: 4 seconds, Moisture flow time delay: 5.5 seconds}. The time delay data is also correlated from the temporal phase coupling data. By performing this correlation operation on all calibrated spatial locations, a multimodal risk distribution dataset is finally established.

[0105] Please see Figure 6 The specific steps of S5 are as follows:

[0106] S501: Extract potential unstable phase regions within the same time window from the multimodal risk distribution dataset, calculate the spatial proximity relationship between each potential unstable phase region, determine the degree of aggregation association between adjacent regions based on coordinate distance and gradient magnitude weight, and establish a set of spatially aggregated and associated regions.

[0107] From the multimodal risk distribution dataset, all records marked as "potentially unstable phase regions" within the same time window (e.g., the past 5 minutes) are extracted. Assuming that monitoring points five, six, and seven are identified as potentially unstable regions within this time window, the spatial proximity between these regions is calculated. The proximity criterion is a preset coordinate distance threshold, determined based on statistical analysis of the core impact range of a hazardous area within a short time in a typical chemical spill diffusion model; in this example, it is 2.5 meters. The Euclidean distances between monitoring points five and six, five and seven, and six and seven are calculated. If the distance between monitoring points five and six is ​​2.0 meters (less than 2.5 meters), then monitoring points five and six are considered adjacent. Subsequently, the degree of aggregation association between adjacent regions is determined based on the coordinate distance and gradient magnitude weight. The gradient magnitude weight is calculated by dividing the concentration gradient magnitude of the region by a baseline concentration gradient (this baseline is 50% of the concentration alarm gradient specified in the safety regulations, set to 5 ppm / m). The concentration gradient at monitoring point 5 is 3.5 ppm / m, with a weight of 3.5 / 5 = 0.7. The concentration gradient at monitoring point 6 is 4.5 ppm / m, with a weight of 4.5 / 5 = 0.9. The degree of aggregation correlation is calculated by adding the weights of the two regions and then dividing by the square of the distance between them, i.e., (0.7 + 0.9) / (2.0²) = 0.4. A aggregation correlation threshold of 0.3 is set. This threshold is determined by analyzing historical leakage accident data to inversely determine the strength of the interaction between regions that lead to chain reactions or risk expansion. Since the calculated correlation degree of 0.4 is greater than the threshold of 0.3, monitoring points 5 and 6 are determined to be aggregated and together form a spatial aggregation correlation region. This judgment is performed on all unstable region pairs to finally establish a set of spatial aggregation correlation regions.

[0108] S502: Based on the multimodal risk distribution dataset, obtain the thermal diffusion time delay and gradient data within the spatially aggregated associated region set, calculate the average thermal diffusion phase difference and average wet flow phase difference of each aggregated region in the spatially aggregated associated region set, and determine the consistency between the phase difference change trend and the gradient change direction within the aggregated region in the spatially aggregated associated region set, and generate the risk assessment results of the aggregated region.

[0109] The process of determining the consistency between the phase difference change trend and the gradient change direction within the spatially aggregated region includes:

[0110] Traverse each aggregation region in the spatial aggregation association region set and extract continuous time series data of thermal diffusion time delay, moisture flow time delay, concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient within the aggregation region;

[0111] Based on the continuous time series, calculate the trend vectors of the thermal diffusion time delay and the wet flow time delay;

[0112] Based on the continuous time series, calculate the combined change direction vectors of temperature gradient and concentration gradient, and the combined change direction vectors of humidity gradient and airflow velocity gradient, respectively.

[0113] Calculate the angle between the trend vector of the thermal diffusion time delay and the combined change direction vector of the temperature gradient and concentration gradient, and compare the angle with a preset consistency angle threshold.

[0114] Simultaneously, the angle between the trend vector of the change in wet flow time delay and the combined change direction vector of humidity gradient and airflow velocity gradient is calculated, and the angle is compared with the consistency angle threshold.

[0115] When the angle calculated by both types is less than the consistency angle threshold, the phase difference change trend is determined to be consistent with the gradient change direction.

[0116] Based on the multimodal risk distribution dataset, thermal diffusion time delay and gradient data are obtained within a spatially aggregated region set (e.g., an aggregated region consisting of monitoring point 5 and monitoring point 6). First, the average thermal diffusion phase difference and average wet flow phase difference of this aggregated region are calculated. Specifically, the thermal diffusion time delay (4 seconds for monitoring point 5, 4.2 seconds for monitoring point 6) and wet flow time delay (5.5 seconds for monitoring point 5, 5.8 seconds for monitoring point 6) are extracted for each monitoring point, and then their average values ​​are calculated. The average thermal diffusion phase difference is (4 + 4.2) / 2 = 4.1 seconds. The average wet flow phase difference is (5.5 + 5.8) / 2 = 5.65 seconds. Next, the consistency between the phase difference change trend and the gradient change direction within the aggregated region is determined. This process requires analysis of continuous time series data. Extract the heat diffusion time delay sequence every 5 seconds from monitoring points 5 and 6 over the past 30 seconds, for example, the sequence [4.0s, 4.0s, 4.1s, 4.1s, 4.2s, 4.3s]. Calculate the slope of this sequence using linear regression, obtaining a positive value indicating an increasing trend. Simultaneously, extract the combined change direction vector of the temperature and concentration gradients within the aggregation region. This vector is obtained by summing the temperature and concentration gradient vectors at each time point and then averaging these sums over the time series. Assume the calculated combined change direction vector is (1.2, -0.3, 0). The consistency criterion is defined as: when the time delay sequence shows a continuous increasing trend, the gradient magnitude sequence also shows a continuous increasing trend. Set a consistency angle threshold of 45 degrees, based on the statistical results from fluid dynamics simulations showing a strong correlation between the increase in time delay and the enhancement of spatial gradient when instability intensifies. When the angle between the changing trends of two types of parameters (e.g., the delay growth trend vector and the gradient growth trend vector) is less than this threshold, they are determined to be in the same direction. This determination result (consistent or inconsistent) is combined with the calculated average phase difference to generate the aggregated regional risk assessment result.

[0117] S503: Compare the average thermal diffusion phase difference and average wet flow phase difference of the risk assessment results of the aggregation area with their respective risk judgment thresholds. When either average phase difference exceeds the corresponding risk judgment threshold, generate a record on the corresponding spatial coordinates of the aggregation area and establish a risk warning result.

[0118] The average thermal diffusion phase difference and average wet flow phase difference in the risk assessment results of the aggregated area are compared with their respective risk judgment thresholds. The risk judgment thresholds are set based on a retrospective analysis of historical safety accident cases. The analysis found that in the critical stage before an actual leak or runaway reaction, the thermal diffusion time delay generally exceeds 4.0 seconds, and the wet flow time delay generally exceeds 5.5 seconds. To allow for warning time, the thermal diffusion risk judgment threshold is set to 4.0 seconds, and the wet flow risk judgment threshold is set to 5.5 seconds. Taking the aggregated area composed of monitoring points five and six as an example, the average thermal diffusion phase difference in its risk assessment results is 4.1 seconds, and the average wet flow phase difference is 5.65 seconds. First, the average thermal diffusion phase difference of 4.1 seconds is compared with the risk judgment threshold of 4.0 seconds. 4.1 seconds exceeds 4.0 seconds, meeting the risk condition. Since the rule is that any average phase difference exceeding the corresponding threshold triggers the warning mechanism, a risk warning record is then activated. A risk warning record is then generated on the spatial coordinates of the aggregated area (taking the coordinates of the center points of monitoring points five and six). The recorded information includes: {Warning Time: Current timestamp, Warning Location: Center coordinates of the aggregated area, Trigger Type: Exceeding thermal diffusion phase difference limit, Trigger Value: 4.1 seconds, Associated Gradient: [Gradient data from monitoring point five and monitoring point six]}. These generated records are aggregated to establish the final risk warning result.

[0119] Please see Figure 7 A chemical laboratory risk early warning system based on multimodal data fusion algorithms includes:

[0120] The data gradient module acquires gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculates the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establishes a multimodal synchronous monitoring dataset.

[0121] The spatial coupling positioning module calculates thermal diffusion coupling parameters and wet flow coupling parameters based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, determines the target spatial location with obvious coupling, and extracts local coupling change data at the target spatial location.

[0122] The phase coupling analysis module analyzes the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupling change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and performs fusion matching to obtain time-series phase coupling data.

[0123] The risk area identification module determines the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on temporal phase coupling data, and integrates them into a multimodal risk distribution dataset.

[0124] The risk aggregation and early warning module performs regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk early warning results.

[0125] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A chemical laboratory risk early warning method based on a multimodal data fusion algorithm, characterized in that, Includes the following steps: S1: Acquire gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculate the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establish a multimodal synchronous monitoring dataset. S2: Based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, calculate the thermal diffusion coupling parameters and the wet flow coupling parameters, determine the target spatial location with obvious coupling, and extract the local coupling change data at the target spatial location; S3: Analyze the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupling change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and perform fusion matching to obtain temporal phase coupling data; S4: Determine the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on the time-series phase coupling data, and integrate them into a multimodal risk distribution dataset; S5: Perform regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk warning results; The specific steps of S2 are as follows: S201: Extract the direction parameters of the concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient in the multimodal synchronous monitoring data set, calculate the angle between the directions of the temperature gradient and the concentration gradient for the same spatial coordinate position to obtain the thermal diffusion coupling parameter, and calculate the angle between the directions of the humidity gradient and the airflow velocity gradient to obtain the moisture flow coupling parameter, and establish a thermal-humidity coupling parameter set. S202: Call the preset thermal diffusion coupling threshold and wet flow coupling threshold, and based on the comparison of the thermal and wet flow coupling parameter set, filter the spatial locations where the thermal diffusion coupling parameter is less than the thermal diffusion coupling threshold and the wet flow coupling parameter is less than the wet flow coupling threshold, as the target spatial locations with obvious coupling; S203: Extract the concentration change rate, temperature change rate, humidity change rate, and airflow velocity change rate corresponding to the target spatial location with obvious coupling from the multimodal synchronous monitoring dataset, and establish local coupling change data; The specific steps for S3 are as follows: S301: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the gas concentration change rate sequence and the temperature change rate sequence at the target spatial location, and obtain the thermal diffusion time delay; S302: Call the target spatial location with obvious coupling in the local coupling change data, calculate the time delay between the humidity change rate sequence and the airflow velocity change rate sequence of the target spatial location, and obtain the wet flow time delay; S303: Invoke the time sequence and spatial coordinates corresponding to the heat diffusion time delay and the wet flow time delay to establish time-phase coupled data; The specific steps for S4 are as follows: S401: Compare the thermal diffusion time delay and the wet flow time delay in the time-series phase coupling data with their respective thermal diffusion thresholds and preset wet flow thresholds to generate phase region calibration results; S402: Based on the spatial locations calibrated in the phase region calibration results, find and extract the concentration gradient data, temperature gradient data, humidity gradient data, and airflow velocity gradient data corresponding to these locations from the multimodal synchronous monitoring dataset, and establish a target region gradient dataset; S403: Based on the common spatial location information in the two data sets, namely the phase region calibration result and the target region gradient dataset, establish a multimodal risk distribution dataset; The specific steps of S5 are as follows: S501: Extract potential unstable phase regions within the same time window from the multimodal risk distribution dataset, calculate the spatial proximity relationship between each potential unstable phase region, determine the degree of aggregation association between adjacent regions based on coordinate distance and gradient magnitude weight, and establish a set of spatially aggregated and associated regions. S502: Based on the multimodal risk distribution dataset, obtain the thermal diffusion time delay and gradient data within the spatially aggregated associated region set, calculate the average thermal diffusion phase difference and average wet flow phase difference of each aggregated region in the spatially aggregated associated region set, determine the consistency between the phase difference change trend and the gradient change direction within the aggregated region in the spatially aggregated associated region set, and generate the aggregated region risk assessment result. S503: Compare the average thermal diffusion phase difference and the average wet flow phase difference of the risk assessment results of the aggregation area with their respective risk judgment thresholds. When any average phase difference exceeds the corresponding risk judgment threshold, generate a record on the corresponding spatial coordinates of the aggregation area and establish a risk warning result.

2. The chemical laboratory risk early warning method based on multimodal data fusion algorithm according to claim 1, characterized in that, The multimodal synchronous monitoring dataset includes concentration gradient distribution, temperature gradient distribution, humidity gradient distribution, and airflow velocity gradient distribution. The local coupled change data specifically refers to the rate of change of concentration, temperature, humidity, and airflow velocity at the target spatial location. The temporal phase coupled data includes thermal diffusion time delay, moisture flow time delay, spatial coordinates, and time markers. The multimodal risk distribution dataset includes information on cooperatively stable phase regions, information on potentially unstable phase regions, and phase difference parameters corresponding to the regions. The risk warning result specifically refers to the spatial location and risk level of the risk.

3. The chemical laboratory risk early warning method based on multimodal data fusion algorithm according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Acquire raw monitoring data collected by gas concentration sensors, temperature sensors, humidity sensors and airflow velocity sensors at designated locations in the chemical laboratory, and perform synchronization alignment based on sensor timestamps. Record the gas concentration, temperature, humidity and airflow velocity at the same time point, associate them with the three-dimensional spatial coordinates marked on the sensor layout map, and generate a monitoring data spatial coordinate mapping set. S102: Based on the monitoring data spatial coordinate mapping set, determine the adjacent monitoring points in the three-dimensional spatial coordinate system, call the coordinate data of the adjacent monitoring points to calculate the Euclidean spatial distance between the two points, and obtain the gas concentration difference, temperature difference, humidity difference and airflow velocity difference of the adjacent monitoring points at the same time stamp. Divide each difference by the corresponding spatial distance to obtain the multidimensional monitoring data spatial gradient value. S103: Call the monitoring data spatial coordinate mapping set and the multi-dimensional monitoring data spatial gradient value, and establish a joint index for gas concentration, temperature, humidity, airflow velocity and corresponding gradient data in a unified coordinate system to obtain a multi-modal synchronous monitoring dataset.

4. The chemical laboratory risk early warning method based on multimodal data fusion algorithm according to claim 1, characterized in that, The process of comparing each thermal diffusion threshold with the preset moisture flow threshold includes: Iterate through each data record in the time-series phase-coupled data and extract the thermal diffusion time delay and wet flow time delay from the data record; The extracted heat diffusion time delay is compared with the heat diffusion threshold, and the extracted wet flow time delay is compared with the preset wet flow threshold. If the thermal diffusion time delay is less than the thermal diffusion threshold and the wet flow time delay is less than the preset wet flow threshold, then the spatial location associated with the data record is marked as the cooperative stable phase region. If the thermal diffusion time delay is greater than or equal to the thermal diffusion threshold, or the wet flow time delay is greater than or equal to the preset wet flow threshold, then the spatial location associated with the data record will be marked as a potentially unstable phase region.

5. The chemical laboratory risk early warning method based on multimodal data fusion algorithm according to claim 1, characterized in that, The process of determining the consistency between the phase difference change trend and the gradient change direction within the spatial aggregation correlation region includes: Traverse each aggregation region in the spatial aggregation association region set and extract continuous time series data of thermal diffusion time delay, moisture flow time delay, concentration gradient, temperature gradient, humidity gradient and airflow velocity gradient within the aggregation region; Based on the continuous time series, calculate the trend vectors of the thermal diffusion time delay and the wet flow time delay; Based on the continuous time series, calculate the combined change direction vectors of temperature gradient and concentration gradient, and the combined change direction vectors of humidity gradient and airflow velocity gradient, respectively. Calculate the angle between the trend vector of the thermal diffusion time delay and the combined change direction vector of the temperature gradient and concentration gradient, and compare the angle with a preset consistency angle threshold. Simultaneously, the angle between the trend vector of the change in wet flow time delay and the combined change direction vector of humidity gradient and airflow velocity gradient is calculated, and the angle is compared with the consistency angle threshold. When the angle calculated by both types is less than the consistency angle threshold, the phase difference change trend is determined to be consistent with the gradient change direction.

6. A chemical laboratory risk early warning system based on a multimodal data fusion algorithm, characterized in that, The system is used to implement the chemical laboratory risk early warning method based on multimodal data fusion algorithm as described in any one of claims 1-5, and the system comprises: The data gradient module acquires gas concentration, temperature, humidity and airflow velocity data at designated locations in the chemical laboratory, calculates the gradient of each type of data between adjacent monitoring points in their respective preset three-dimensional spaces, and establishes a multimodal synchronous monitoring dataset. The spatial coupling positioning module calculates thermal diffusion coupling parameters and moisture flow coupling parameters based on the gradient direction of each type of data in the multimodal synchronous monitoring dataset, determines the target spatial location with obvious coupling, and extracts local coupling change data at the target spatial location. The phase coupling analysis module analyzes the thermal diffusion time delay of gas concentration and temperature changes at the target spatial location in the local coupling change data, as well as the wet flow time delay of humidity changes and airflow velocity changes, and performs fusion matching to obtain time-series phase coupling data. The risk area identification module determines the co-stable or potentially unstable phase regions of thermal diffusion and moisture flow based on the time-series phase coupling data, and integrates them into a multimodal risk distribution dataset. The risk aggregation and early warning module performs regional aggregation analysis along the laboratory spatial coordinates on the multimodal risk distribution dataset to generate risk early warning results.

Images