Factory safety intelligent management and control method based on heterogeneous multi-system cross service fusion technology

Abstract

The invention relates to the technical field of industrial safety, in particular to a factory safety intelligent management and control method based on a heterogeneous multi-system cross service fusion technology, and the method comprises the steps: collecting factory multi-source heterogeneous data, and carrying out the time-space alignment and fusion; constructing a risk assessment model based on a deep learning algorithm, and performing quantitative risk analysis on personnel behaviors, environmental parameters and equipment states; a dynamic grading early warning mechanism is established, and warning and automatic handling are achieved according to the risk grade; introducing risk trend prediction, and analyzing a risk evolution trend; and a safe closed-loop optimization mechanism is constructed, and self-adaptive adjustment of the management and control strategy is realized. According to the method, a full-process and dynamically-optimized intelligent safety management and control scheme is constructed, and comprehensive perception, quick response and continuous optimization of factory safety management are realized.

Description

Technical Field

[0001] This application relates to the field of industrial safety technologies, and specifically to a method for intelligent control and management of plant area safety based on heterogeneous multi-system cross-service fusion technology. Background Art

[0002] Multiple heterogeneous business systems will be deployed in the plant area, including industrial Internet of Things systems, video surveillance systems, access control systems, and environmental monitoring systems, etc. Due to the inconsistent data formats and interface standards of each system, and the data is scattered and stored in their respective business platforms, lacking an effective data fusion mechanism, the data value fails to be fully explored and utilized, forming an information island for safety management.

[0003] The safety risks in industrial parks are characterized by complexity and diversity. Factors such as personnel behavior, environmental changes, and equipment failures are intertwined and affect each other. A single-dimensional risk assessment method is difficult to comprehensively reflect the safety status of the park. And currently, most park safety warning systems lack quantitative risk assessment means, and it is difficult to ensure accuracy and real-time performance, and situations such as missed alarms and false alarms are likely to occur, affecting the warning effect.

[0004] In view of this, this application proposes a method for intelligent control and management of plant area safety based on heterogeneous multi-system cross-service fusion technology. Summary of the Invention

[0005] To achieve the above object, the present invention provides a method for intelligent control and management of plant area safety based on heterogeneous multi-system cross-service fusion technology. The specific technical solutions are as follows:

[0006] A method for intelligent control and management of plant area safety based on heterogeneous multi-system cross-service fusion technology includes:

[0007] Collect plant area data, align the collected plant area data from the time and space dimensions, fuse the aligned plant area data, use it as plant area event fusion data and extract data features, and use the plant area event fusion data and the extracted feature data as plant area event feature fusion data;

[0008] Based on deep learning algorithms, construct a risk assessment model, and conduct risk assessments on personnel behavior, environmental parameters, and equipment status of the plant area according to the plant area event feature fusion data;

[0009] Establish a dynamic hierarchical warning mechanism, classify the risk levels of events according to the risk assessment results, and send warning information to a preset receiving end and trigger an automatic disposal process based on the dynamic hierarchical warning mechanism;

[0010] Construct a plant area risk trend prediction mechanism, and analyze the evolution trend of plant area risks on the basis of risk assessment, in combination with the plant area event fusion data and the dynamic hierarchical warning results;

[0011] Build a closed-loop optimization mechanism for plant safety to adaptively adjust and optimize the safety control of the plant.

[0012] Preferably, the plant data includes: industrial Internet of Things (IIoT) terminal data, monitoring video data, access control data, and environmental data; the IIoT terminal data includes: equipment status, process parameters, energy consumption parameters, and fault codes; the monitoring video data is the target detection result in the monitoring video; the access control data includes: the switch status of the access control point, access card swiping records, and access control exception events; the environmental data includes temperature, humidity, gas concentration, and noise.

[0013] Preferably, define a unified spatio-temporal reference system, and perform spatio-temporal alignment for different types of collected data respectively; after completing the spatio-temporal alignment, fuse the data from different sources according to the spatio-temporal position to obtain a unified representation of plant event fusion data.

[0014] Extract features from the plant event fusion data, and use the plant event fusion data and the extracted feature data as the plant event feature fusion data.

[0015] Preferably, divide the plant safety risk assessment problem into a multi-classification task, that is, according to the plant event feature fusion data, determine the risk coefficient of the event; build a risk assessment model to perform risk discrimination on the plant event fusion data.

[0016] Preferably, build a risk assessment model based on a multi-modal deep learning algorithm, and the multi-modal deep learning algorithm includes: a data preprocessing layer, a feature learning layer, a modal fusion layer, and a classification output layer.

[0017] Train the multi-modal deep learning algorithm through a data set, deploy the trained risk assessment model to the online environment, and perform risk discrimination on the real-time collected plant event feature fusion data.

[0018] Preferably, based on the risk assessment model, perform risk discrimination on the collected plant event feature fusion data to obtain the risk coefficient of each event.

[0019] Set a risk probability threshold, and divide the risk level by comparing the risk coefficient and the risk probability threshold.

[0020] Preferably, introduce a dynamic threshold optimization mechanism, define a sliding time window, and record all events and their risk assessment results in the past time.

[0021] For each risk level, count the number of events that truly belong to the divided level and the number of events with a predicted probability greater than the current threshold in the assessment results; calculate the warning precision rate and recall rate of the level.

[0022] Adjust the precision and recall rate, and by maximizing the threshold, achieve a balance between the precision and recall rate of the early warning results.

[0023] Preferably, count the number of plant events and the average probability of each risk level, use the ARIMA model in time series analysis to predict the risk trend, and predict the future risk coefficient trend of the plant according to the ARIMA model.

[0024] Preferably, based on the historical risk assessment results, early warning information, and actual accident occurrence situations, conduct regular retrospective analysis on the risk assessment model, and use the incremental learning method to optimize the multi-modal deep learning model;

[0025] Combined with the dynamic threshold optimization mechanism, monitor the changes in the precision and recall rate of early warning, and dynamically adjust the risk level division threshold according to the effectiveness feedback of historical early warnings to optimize the early warning strategy.

[0026] The plant safety intelligent control system based on the heterogeneous multi-system cross-business fusion technology, which is used to implement the plant safety intelligent control method based on the heterogeneous multi-system cross-business fusion technology, includes: a data acquisition module, a risk assessment module, a hierarchical early warning module, a risk trend prediction module, and a closed-loop optimization module;

[0027] The data acquisition module is used to collect plant data, align the collected plant data from the time and space dimensions, fuse the aligned plant data, use it as the plant event fusion data and extract data features, and use the plant event fusion data and the extracted feature data as the plant event feature fusion data;

[0028] The risk assessment module, based on the deep learning algorithm, constructs a risk assessment model, and conducts risk assessments on the personnel behavior, environmental parameters, and equipment status of the plant according to the plant event feature fusion data;

[0029] The hierarchical early warning module is used to establish a dynamic hierarchical early warning mechanism, classify the events according to the risk assessment results, and send warning information to the preset receiving end and trigger the automatic disposal process based on the dynamic hierarchical early warning mechanism;

[0030] The risk trend prediction module is used to construct a plant risk trend prediction mechanism, and on the basis of risk assessment, combine the plant event fusion data and the dynamic hierarchical early warning results to analyze the evolution trend of the plant risk;

[0031] The closed-loop optimization module is used to construct a plant safety closed-loop optimization mechanism to adaptively adjust and optimize the plant safety control.

[0032] Advantages of the present invention: By performing spatio-temporal alignment and fusion on heterogeneous data, this application eliminates data islands and realizes the interconnection, interoperability, and comprehensive utilization of data.

[0033] By constructing a deep learning algorithm, this application fully explores the complex correlation features contained in the fused data and realizes the quantitative assessment of multi-dimensional risk factors such as personnel, environment, and equipment.

[0034] This application realizes hierarchical early warning according to the risk assessment results, making the early warning more accurate and timely, reducing false negatives and false positives; at the same time, it triggers automatic disposal, improving the timeliness of early warning response.

[0035] By analyzing historical data and assessment results, this application predicts the evolution trend of risks, transforming safety management from passive response to proactive prevention.

[0036] By continuously monitoring and feedback on the control process, this application dynamically optimizes the risk assessment model and early warning strategy, improving the accuracy and effectiveness of safety control and realizing the adaptive adjustment of the strategy.

[0037] This application integrates data collection, risk assessment, early warning response, trend analysis, and strategy optimization into an intelligent safety control closed-loop, realizing the deep integration of data, algorithms, and services, significantly improving the intelligent and refined level of plant safety management, and having important significance for ensuring the stable operation of the park and avoiding safety accidents. Brief Description of the Drawings

[0038] Figure 1 It is a flowchart of the plant safety intelligent control method based on the heterogeneous multi-system cross-service fusion technology provided by the present invention;

[0039] Figure 2 It is a schematic diagram of the data collection and fusion process provided by the present invention;

[0040] Figure 3 It is a schematic diagram of the risk assessment model construction process provided by the present invention;

[0041] Figure 4 It is a schematic diagram of the dynamic hierarchical early warning process provided by the present invention;

[0042] Figure 5 It is a schematic diagram of the risk trend prediction process provided by the present invention;

[0043] Figure 6 It is a structure diagram of the plant safety intelligent control system based on the heterogeneous multi-system cross-service fusion technology provided by the present invention. Detailed Embodiments

[0044] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following detailed description of the specific embodiments of the present invention will be given in conjunction with the accompanying drawings of the specification.

[0045] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0046] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive of other embodiments.

[0047] Embodiment 1

[0048] Refer to Figures 1 to 5 , the first embodiment of the present invention provides a method for intelligent control and management of plant safety based on heterogeneous multi-system cross-service fusion technology.

[0049] Step 1: Collect plant data, align the collected plant data from the time and space dimensions, fuse the aligned plant data, use it as plant event fusion data and extract data features, and use the plant event fusion data and the extracted feature data as plant event feature fusion data.

[0050] Define the data collected by the i-th Internet of Things terminal at time t as an I-dimensional vector where I is the number of data types collected by the terminal.

[0051] The data types collected by the Internet of Things terminal include: equipment status s i (t) ∈ {0, 1}, process parameters energy consumption parameters and fault codes where M p , M e , M c are the dimensions of each data type respectively, is a sparse matrix, then the data of the i-th terminal can be expressed as The data dimension is M = 1 + M p + M e + M c .

[0052] There are K monitoring video cameras, and the original video frame collected by the k-th camera at time t is Where H, W, and G are the height, width, and number of channels of the image, respectively.

[0053] Perform object detection on the original video frame, and extract L objects of interest (such as people, vehicles, equipment, etc.). The l-th object is represented by the bounding box vector where (x l , y l ) are the center coordinates of the bounding box, and w l and h l are the width and height of the box.

[0054] Represent the object detection result as where L k is the number of objects detected by the k-th camera at time t.

[0055] Suppose there are D access control controllers, and define the state of the d-th controller at time t as where N d is the number of access control points managed by this controller, represents the switch state of the n-th access control point.

[0056] The access control card swiping record is defined as a triple r j (t) = <u j , d j , t>, where u j is the swiping user ID, and d j is the controller number of the swiping.

[0057] The access control abnormal event (such as the door not closed, illegal intrusion, etc.) is defined as e q (t) = <d q , c q , t>, where d q is the controller number where the event occurs, and c q is the event type code.

[0058] Suppose there are E environmental sensors, and the data collected by the e-th sensor at time t is The data of different types of sensors include: temperature tmp(t), humidity hmd(t), gas concentration gas(t), and noise db(t).

[0059] The data collected by the sensors is uniformly represented as an E×4-dimensional vector:

[0060]

[0061] Since the multi-source heterogeneous data collected may be inconsistent in the time and space dimensions, spatio-temporal alignment processing is required.

[0062] Define a unified spatio-temporal reference system. Let the starting point of the time axis be \(t_0\) and the ending point be \(t\). end The time granularity is \(\Delta t\); the origin of the spatial coordinate system is \(O(0, 0, 0)\).

[0063] For each data acquisition device, where the data acquisition device is any device that acquires industrial Internet of Things terminal data, surveillance video data, access control data, and environmental data, define the spatio-temporal acquisition domain of the data acquisition device as \(D\). s (t, x, y, z), representing the data acquired at time \(t\) and spatial location \((x, y, z)\).

[0064] For different data types, different spatio-temporal alignment methods are adopted:

[0065] For Internet of Things terminal data Since the terminal location is fixed, only alignment in the time dimension is required; map the acquisition time series to the unified time axis to obtain the aligned data [t0, t end .

[0066] For surveillance video data For each detection target Convert its bounding box coordinates to the spatial coordinate system, denoted as Then align the target data of different cameras in the time dimension to obtain

[0067] For access control data Map each access control controller and the event occurrence location to the spatial coordinate system, denoted as The alignment method in the time dimension is the same as that of Internet of Things terminals.

[0068] For environmental data Map the location of each sensor to the spatial coordinate system, denoted as The alignment method in the time dimension is the same as that of Internet of Things terminals.

[0069] After completing spatio-temporal alignment, fuse the data from different sources according to spatio-temporal positions to obtain a unified representation of factory area event fusion data.

[0070] Define a factory area event as a five-tuple where \((t, x, y, z)\) are the spatio-temporal coordinates of the event, is the event feature vector, containing multi-source heterogeneous data of Internet of Things terminal data, surveillance video data, access control data, and environmental data.

[0071] For each spatio-temporal acquisition domain \(D\). s (t, x, y, z), extract the heterogeneous data features therein to form a local event representation The local events at the same time and space location are fused to obtain a complete plant event representation: Among them, ∪ represents the feature cascade or fusion operation, which can be implemented by vector concatenation, weighted average or maximum pooling.

[0072] Extract features from the plant event fusion data to obtain higher-level feature representation for subsequent risk analysis and early warning; define a set of feature extraction functions {φ1, φ2, ..., φ K}, encode different attributes of plant events; exemplary, time series feature extraction function: extract the statistics of event duration and frequency; spatial feature extraction function: extract the geometric quantities of spatial distribution and clustering density of events; after applying the feature extraction function, the event feature vector is obtained: Where N1 is the feature vector dimension.

[0073] The plant event fusion data and the extracted feature data after time-space alignment and fusion are expressed as plant event feature fusion data {W}:

[0074]

[0075] Where M is the total number of events, are the original features and extracted features of the mth event respectively.

[0076] See also Figure 2 , which is a schematic diagram of the data collection and fusion process.

[0077] This step eliminates data silos by aligning and fusing heterogeneous data in time and space, realizes data interconnection and comprehensive utilization, provides comprehensive and accurate data support for subsequent risk assessment, early warning and optimization, and improves data value density.

[0078] Step 2: Based on the deep learning algorithm, a risk assessment model is constructed to conduct risk assessment of personnel behavior, environmental parameters, and equipment status in the factory area based on the fusion data of factory event characteristics.

[0079] Defining risk level sets where r C is the risk level, C is the number of risk levels; each event W m The corresponding risk factor

[0080] By training the risk assessment model f θ , predict the risk coefficient of the new plant event feature fusion data; where θ is the model parameter.

[0081] Adopt a multi-modal deep learning algorithm as the risk assessment model to make full use of the complementary information of heterogeneous data; the multi-modal deep learning algorithm includes: a data preprocessing layer, a feature learning layer, a modal fusion layer, and a classification output layer;

[0082] The data preprocessing layer is used to perform standardization and normalization operations on the original event features and map them to a unified feature space; denote the processed features as

[0083] For the extracted features different encoding methods are adopted according to the feature type (numerical or categorical), using Embedding encoding or One-hot encoding; denote the encoded features as

[0084] The feature learning layer includes: using independent sub-networks to perform feature learning on data of different modalities.

[0085] For the structured data of Internet of Things terminal data and environmental data, a multi-layer perceptron (MLP) model is used to perform non-linear transformation: are the structured data features of Internet of Things terminal data and environmental data.

[0086] For the surveillance video data, a 3D convolutional neural network (3D-CNN) model is used to extract spatio-temporal features: are the surveillance video data features.

[0087] For the discrete data of access card swiping and access abnormal events, a recurrent neural network (RNN) is used to capture the temporal dependencies: are the discrete data features of access card swiping and access abnormal events.

[0088] For the extracted features the attention mechanism is adopted to automatically learn the importance weights of different features: are the features extracted by the attention mechanism; where is the attention parameter, d s , d v , d d , d f is the output dimension of the sub-network.

[0089] The modal fusion layer is used to fuse the features learned from different modalities to obtain the comprehensive representation of the event: where Fu(·) can be any one of the concatenation, summation, and attention fusion functions.

[0090] The classification output layer includes: based on event representation Predict its risk coefficient:

[0091]

[0092] where is the classifier parameter, Δ C-1 is a C-1 dimensional simplex, representing the predicted probability distribution, is the predicted risk coefficient, represents event W m is the predicted probability that it belongs to the risk coefficient of the c-th category, represents selecting the category with the highest probability.

[0093] Obtain the factory historical data set, and manually annotate the historical data set. Based on the annotated historical event data set Adopt the cross-entropy loss function and the backpropagation algorithm to train the risk assessment model.

[0094] Use the Adam optimizer and mini-batch gradient descent method to minimize the loss function and update the risk assessment model parameter θ until the model converges or reaches the preset number of iterations.

[0095] Evaluate the risk coefficient of the real-time collected factory area event feature fusion data through the trained risk assessment model.

[0096] See Figure 3 , which is the schematic diagram of the risk assessment model construction process in this step.

[0097] In this step, the deep learning algorithm is used to fully mine the complex correlation features contained in the fusion data, realizing the quantitative assessment of multi-dimensional risk factors such as personnel, environment, and equipment, overcoming the limitations of traditional methods, and providing comprehensive and accurate risk discrimination results.

[0098] Step 3: Establish a dynamic hierarchical early warning mechanism. According to the risk assessment results, classify the risk levels of events, and send warning messages to the preset receiving end and trigger the automatic disposal process based on the dynamic hierarchical early warning mechanism.

[0099] Based on the risk assessment model f θ , evaluate the risk coefficient of the collected factory area event feature fusion data {W}, and obtain the risk coefficient of each event in the factory area event feature fusion data Set risk probability thresholds τ1, τ2,..., τ C , satisfying 0 < τ1 < τ2 <... < τ C-1 < τ C ≤ 1; then event W mRisk level L m Is determined as:

[0100]

[0101] If event W m Risk coefficient of Belongs to a certain risk threshold range, then event W m Risk level is classified into this risk level, and the risk probability threshold {τ c} can be set according to requirements.

[0102] To adapt to the changes in the scenario and improve the sensitivity of early warning, a dynamic threshold optimization mechanism is introduced; define a sliding time window Δht to record all events and their risk assessment results within the past Δht time

[0103] For each risk level r c , count the number of events N that truly belong to this level in c and the number of events M1 with predicted probability greater than the current threshold τ c ; calculate the early warning precision P c and recall rate R c of level r c :

[0104]

[0105] Wherein is an indicator function, which takes the value of 1 when the condition is met and 0 otherwise, and M c is the number of events in the evaluation result .

[0106] Weigh the precision and recall rate through the F 1c value:

[0107]

[0108] Adjust the threshold τ 1c by maximizing F c to make the early warning result achieve a balance between precision and recall rate:

[0109]

[0110] Recalculate and update the threshold at regular intervals to achieve the adaptive optimization of the early warning strategy.

[0111] Exemplarily, according to the risk level L m of the event, trigger the corresponding early warning and handling processes:

[0112] For low-risk events (L m = r1), generate a warning message and send it to on-site management personnel to prompt them to pay attention to monitoring;

[0113] For medium-risk events (L m = r2), in addition to the warning message, automatically trigger the video surveillance system to capture and record videos of relevant areas, and notify management personnel to arrive at the scene for investigation;

[0114] For high-risk events (L m ≥ r3), on the basis of warnings and captures, further take the following measures:

[0115] Automatically trigger the nearby access control system to block the area and control personnel entry and exit; send a shutdown command to relevant production equipment to prevent endangerment to personal and property safety; issue evacuation or risk avoidance instructions through the emergency broadcast system to guide personnel to evacuate safely; notify the emergency rescue team to respond quickly and arrive at the scene for disposal; the content of the warning message includes: the time and location of the event, the personnel / equipment / areas involved, the risk level, and disposal requirements, etc., and is sent to relevant personnel and systems in the form of text, voice, images, etc.

[0116] After issuing a warning, track the subsequent status of the event in real time until the event is properly handled;

[0117] Define the set of event statuses including "warning", "being handled", and "resolved"; each event W m in addition to the risk level attribute L m also has a status attribute initialized to S1;

[0118] Exemplarily, after a warning is issued, relevant personnel and systems report the progress of handling and update the event status; when management personnel arrive at the scene, update S i to "being handled"; when the event is controlled and handled, update S i to "resolved"; for events that have been in a certain intermediate state for a long time, urge relevant personnel to handle them as soon as possible, and if necessary, increase the warning level.

[0119] See Figure 4 for the process schematic diagram of the dynamic hierarchical warning in this step.

[0120] This step realizes hierarchical warning according to the risk assessment results, making the warning more accurate and timely, reducing missed reports and false alarms; at the same time, triggering automatic disposal, improving the timeliness of warning response, reducing the possibility of accidents, and ensuring the safe and stable operation of the park.

[0121] Step 4: Build a risk trend prediction mechanism for the factory area. Based on the risk assessment, combine the integrated data of factory area events and the results of dynamic classification early warning to analyze the evolution trend of the factory area risks.

[0122] Count the number of factory area events N t , and the proportion w c of events for each risk level r tc :

[0123]

[0124] where is an indicator function; the time series of the risk distribution and calculate the average probability of each risk level within this time period:

[0125]

[0126] Obtain the time series of the risk probabilities The risk time series dataset is expressed as: where T is the total number of time steps; the risk time series dataset contains the dynamic changes of the factory area risk distribution and risk probabilities.

[0127] Use the ARIMA model in time series analysis to conduct risk trend prediction. The ARIMA model is an autoregressive integrated moving average model. The ARIMA model combines the characteristics of three models: autoregressive (AR), differencing (I), and moving average (MA), and can better fit and predict non-stationary time series.

[0128] Let be the time series of the factory area risk levels, where μ mt represents the risk coefficient of event W m at the t-th moment; the ARIMA(A, g, B) model can be expressed as:

[0129]

[0130] where A is the order of the autoregressive term, g is the order of differencing, B is the order of the moving average term, Ω is the lag operator, defined as Ω a μ mt = μ mt-a ; δ a is the autoregressive coefficient, σ b is the moving average coefficient, and ε t is the random disturbance term, usually assumed to follow a normal distribution with a mean of 0.

[0131] The construction and application process of the ARIMA model is as follows:

[0132] Test the stationarity of the ARIMA model. Use the ADF test or the KPSS test to determine whether the original risk sequence {μ mt} is stationary. If it is not stationary, perform differencing operations until it becomes stationary. Denote the differenced sequence as {μ′ mt}.

[0133] Determine the order of the ARIMA model. Observe the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of the sequence {μ′ mt}. Based on the trailing and truncation situations of the graphs, preliminarily determine the values of A and B. Score models of different orders and select the (A, g, B) combination with the best score.

[0134] Estimate the parameters of the ARIMA model. Use the maximum likelihood estimation method or the conditional least squares method to estimate the model parameters δ a and σ b .

[0135] Diagnose the ARIMA model. Conduct residual analysis on the fitted model to test whether the residual sequence is white noise, that is, whether there is significant autocorrelation. If the model passes the diagnosis, it can be used for prediction; otherwise, the order of the model needs to be reselected.

[0136] Use the fitted ARIMA(A, g, B) model to predict the risk coefficients at the next h time points:

[0137]

[0138] where and are the estimated model parameters, is the prediction error, usually taken as 0;

[0139] According to the future risk levels predicted by the ARIMA model Combine with the pre-set risk level thresholds to generate corresponding risk warning information.

[0140] Exemplarily, set the representation form when the risk coefficient has an upward trend: If then it indicates that the risk coefficient in the factory area will increase within the next h time points, and preventive measures need to be taken in advance to strengthen monitoring and control; if then it indicates that the risk coefficient will increase rapidly within the next h time points, and the emergency plan needs to be activated immediately to make emergency preparations.

[0141] The warning information can be pushed to relevant management personnel and decision-makers in the forms of charts, texts, voices, etc., to help them timely understand the risk trends, optimize resource allocation, and improve the risk prevention and emergency response capabilities.

[0142] See Figure 5, which is a schematic diagram of the risk trend prediction process for this step.

[0143] Based on the ARIMA time series model, this step constructs a technical solution for predicting and warning the risk trend in the factory area, expands the time dimension of risk assessment and warning, enabling it to make forward-looking judgments and responses to the risk evolution in the future for a period of time, which has important practical significance.

[0144] Step 5: Construct an optimized mechanism for the closed-loop safety of the factory area to make adaptive adjustments and optimizations to the safety control of the factory area.

[0145] Based on the historical risk assessment results, warning information, and actual accident occurrence situations, conduct regular retrospective analysis on the risk assessment model, and use the incremental learning method to optimize the multi-modal deep learning model, enabling the multi-modal deep learning model to continuously adapt to the dynamic changes in the factory area environment and improve the accuracy of risk discrimination.

[0146] Combined with the dynamic threshold optimization mechanism, monitor the changes in the precision rate and recall rate of the monitoring and warning, and through the reinforcement learning method, dynamically adjust the risk level classification threshold according to the effectiveness feedback of historical warnings, optimize the warning strategy, and ensure the best balance between the accuracy and coverage of the warning results.

[0147] On the basis of risk assessment and trend prediction, further construct an intelligent decision-making mechanism to achieve intelligent and automated disposal of factory area safety incidents; classify factory area incidents according to the risk level discrimination results, such as personnel violations, equipment abnormalities, and environmental exceedances, and combine with the factory area safety management specifications to automatically match the best disposal plan, such as warning reminders, human-machine collaborative disposal, or automatic control strategies, etc.

[0148] For high-risk events, an automated emergency disposal process can be triggered, such as automatically cutting off the power supply, closing the valve, starting the fire protection system, etc., while low-risk events can be intervened through intelligent prompts or management personnel review to reduce the interference caused by false alarms.

[0149] Through the above optimization mechanism, a closed-loop safety control system of "data collection → data fusion → risk assessment → warning trigger → trend prediction → intelligent decision-making → automatic disposal → feedback optimization" is finally formed.

[0150] The closed-loop safety control system can continuously learn and optimize during the factory area safety management process, realize the intelligence, automation, and continuous improvement of safety management, thereby improving the accuracy and response speed of factory area safety management and ensuring the stable operation of the factory area.

[0151] Example 2

[0152] Refer to Figure 6, which is the second embodiment of the present invention, provides a factory area safety intelligent control system based on heterogeneous multi-system cross-service fusion technology.

[0153] The system includes: a data acquisition module, a risk assessment module, a hierarchical early warning module, a risk trend prediction module, and a closed-loop optimization module.

[0154] The data acquisition module is used to collect factory area data, align the collected factory area data from the time and space dimensions, fuse the aligned factory area data, use it as factory area event fusion data and extract data features, and use the factory area event fusion data and the extracted feature data as factory area event feature fusion data.

[0155] The risk assessment module constructs a risk assessment model based on a deep learning algorithm, and conducts risk assessments on the personnel behavior, environmental parameters, and equipment status of the factory area according to the factory area event feature fusion data.

[0156] The hierarchical early warning module is used to establish a dynamic hierarchical early warning mechanism, classify the risk levels of events according to the risk assessment results, and send warning messages to the preset receiving end and trigger an automatic handling process based on the dynamic hierarchical early warning mechanism.

[0157] The risk trend prediction module is used to construct a factory area risk trend prediction mechanism, and analyze the evolution trend of the factory area risk on the basis of risk assessment, in combination with the factory area event fusion data and the dynamic hierarchical early warning results.

[0158] The closed-loop optimization module is used to construct a factory area safety closed-loop optimization mechanism to adaptively adjust and optimize the factory area safety control.

[0159] In the above embodiments of the present application, the descriptions of each embodiment have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0160] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the process Figure 1 in one process or multiple processes and / or boxes Figure 1 the functions specified in one box or multiple boxes.

[0161] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0162] The embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above specific embodiments. The above specific embodiments are only illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make changes, modifications, substitutions, and variations to the above embodiments without departing from the purpose of the present invention and the scope protected by the claims. These all fall within the protection scope of the present invention.

Claims

1. A factory area safety intelligent management and control method based on heterogeneous multi-system cross-business fusion technology, characterized in that: include: Collect plant data, align the collected plant data from the time and space dimensions, fuse the aligned plant data as plant event fusion data and extract data features, and use the plant event fusion data and the extracted feature data as plant event feature fusion data; Based on deep learning algorithms, a risk assessment model is built to conduct risk assessment of personnel behavior, environmental parameters, and equipment status in the factory area based on the fusion data of factory event characteristics; Establish a dynamic graded warning mechanism to classify events into risk levels based on risk assessment results, send warning information to the preset receiving end based on the dynamic graded warning mechanism and trigger the automatic handling process; Establish a risk trend prediction mechanism for the plant area. Based on risk assessment, combine the fusion data of plant area events with the dynamic classification warning results to analyze the evolution trend of plant area risks. Build a closed-loop optimization mechanism for factory safety and make adaptive adjustments and optimizations to factory safety management and control.

2. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 1 is characterized in that: The factory data includes: industrial Internet of Things terminal data, monitoring video data, access control data and environmental data; the industrial Internet of Things terminal data includes: equipment status, process parameters, energy consumption parameters and fault codes; the monitoring video data is the target detection result in the monitoring video; the access control data includes: the switch status of the access control point, access control card swiping records and access control abnormal events; the environmental data includes temperature, humidity, gas concentration and noise.

3. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 2 is characterized in that: Define a unified spatiotemporal reference system and perform spatiotemporal alignment for different data types collected; After completing the time-space alignment, the data from different sources are fused according to the time-space position to obtain a unified fusion data representation of plant events; Feature extraction is performed on the factory event fusion data, and the factory event fusion data and the extracted feature data are used as factory event feature fusion data.

4. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 3 is characterized in that: The factory safety risk assessment problem is divided into a multi-classification task, that is, the risk coefficient of the event is determined based on the fused data of the factory event characteristics; a risk assessment model is constructed to perform risk determination on the fused data of the factory event characteristics.

5. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 4 is characterized in that: Constructing a risk assessment model based on a multimodal deep learning algorithm, which includes: a data preprocessing layer, a feature learning layer, a modality fusion layer, and a classification output layer; The multimodal deep learning algorithm is trained through the data set, the trained risk assessment model is deployed to the online environment, and the risk identification is performed on the collected factory event feature fusion data.

6. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 5 is characterized in that: Based on the risk assessment model, the collected plant event feature fusion data is used to identify risks and obtain the risk coefficient of each event; Set the risk probability threshold and divide the risk level by comparing the risk coefficient and the risk probability threshold.

7. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 6 is characterized in that: Introduce a dynamic threshold optimization mechanism, define a sliding time window, and record all events in the past and their risk assessment results; For each risk level, count the number of events that actually belong to the level in the assessment results and the number of events whose predicted probability is greater than the current threshold; Calculate the level of warning precision and recall: Adjust the precision and recall rate, and maximize the threshold so that the warning results achieve a balance between precision and recall rate.

8. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 7 is characterized in that: The number of plant events and the average probability of each risk level are counted, and the ARIMA model in time series analysis is used to predict risk trends. The future trend of the plant's risk coefficient is predicted based on the ARIMA model.

9. The method for intelligent factory safety management and control based on heterogeneous multi-system cross-business fusion technology according to claim 8 is characterized in that: Based on historical risk assessment results, early warning information and actual accident situations, the risk assessment model is regularly retrospectively analyzed, and the multimodal deep learning model is optimized using incremental learning methods; Combined with the dynamic threshold optimization mechanism, monitor the changes in warning accuracy and recall rate, dynamically adjust the risk level classification threshold based on historical warning effectiveness feedback, and optimize the warning strategy.

10. A factory security intelligent management and control system based on heterogeneous multi-system cross-business fusion technology, which is used to implement the factory security intelligent management and control method based on heterogeneous multi-system cross-business fusion technology as described in any one of claims 1 to 9, characterized in that: include: Data collection module, risk assessment module, graded warning module, risk trend prediction module and closed-loop optimization module; The data collection module is used to collect factory data, align the collected factory data from the time and space dimensions, fuse the aligned factory data as factory event fusion data and extract data features, and use the factory event fusion data and the extracted feature data as factory event feature fusion data; The risk assessment module builds a risk assessment model based on a deep learning algorithm, and conducts risk assessment on personnel behavior, environmental parameters, and equipment status in the factory area according to the fusion data of factory area event characteristics; The hierarchical warning module is used to establish a dynamic hierarchical warning mechanism, classify the risk level of the event according to the risk assessment result, send warning information to the preset receiving end based on the dynamic hierarchical warning mechanism and trigger the automatic disposal process; The risk trend prediction module is used to build a plant risk trend prediction mechanism, and analyze the evolution trend of plant risk based on risk assessment, combined with plant event fusion data and dynamic graded warning results; The closed-loop optimization module is used to construct a closed-loop optimization mechanism for factory safety and to adaptively adjust and optimize factory safety management and control.

Images