Hazardous chemical substance safety production management and control system

Abstract

The invention relates to the technical field of hazardous chemical substance production, and provides a hazardous chemical substance safety production management and control system, in the aspect of data acquisition, all-around acquisition of equipment operation, environment and personnel operation data is realized, a high-precision sensor ensures data accuracy and far exceeds traditional finite dimension and precision acquisition, and during data analysis and processing, edge calculation is combined with deep learning, so that the safety production management and control of hazardous chemical substances is realized. A risk assessment model is dynamically adjusted, the limitation of a traditional fixed model is broken through, production change and risk early warning can be adapted in real time, risk levels are dynamically adjusted, multi-element early warning and visual display are achieved, compared with single early warning and simple display, risk perception is more timely and accurate, accurate reasons are found through a unique accident analysis model, and the risk assessment efficiency is improved. According to the method, the production management and operation improvement can be promoted, each module of the system can be optimized, a closed-loop lifting mechanism is formed, the safety production management and control level of the hazardous chemical substances is comprehensively improved from data acquisition, analysis, early warning and optimization, the safety risk is greatly reduced, and the production safety and stability are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of hazardous chemical production, and specifically to a safety production control system for hazardous chemicals. Background Art

[0002] In the field of hazardous chemical production safety, traditional control systems face many severe challenges and urgently need innovative technologies to solve them. In terms of data collection in the existing technology, it is often limited to some operating parameters of key equipment, ignoring environmental and personnel operation data, and the sensor accuracy is poor, unable to accurately capture the equipment status and environmental changes, resulting in insufficient integrity and accuracy of the data.

[0003] In the data analysis and processing link, it mostly relies on fixed models and is difficult to fit the complex and changeable characteristics of hazardous chemical production. When production conditions, environmental factors or personnel operations are dynamically adjusted, it is unable to update the risk assessment results in a timely and accurate manner, seriously affecting the prediction and prevention of potential risks. The risk warning mechanism also has shortcomings. The risk level division lacks flexibility and is mostly set statically, unable to change according to the real-time risk situation in a timely manner. The warning method is single, the information transmission is incomplete, and the visual presentation is simple and crude, making it difficult for management personnel to quickly and comprehensively grasp the risk situation. The accident cause analysis is mostly based on experience or simple data analysis, lacking depth and scientificity. In terms of system optimization, it is difficult to effectively adjust and improve modules such as data collection and analysis according to accident feedback and historical data, and a closed-loop optimization system cannot be formed.

[0004] Therefore, a safety production control system for hazardous chemicals is proposed, aiming to comprehensively improve the safety production control level of hazardous chemicals and ensure production safety and stability. Summary of the Invention

[0005] Technical Problems to be Solved

[0006] In view of the deficiencies of the existing technology, the present invention provides a safety production control system for hazardous chemicals.

[0007] Technical Solutions

[0008] To achieve the above solution purpose, the present invention provides the following technical solutions: A safety production control system for hazardous chemicals, including the following:

[0009] S1: A multi-source data collection module, used to comprehensively collect data during the production process of hazardous chemicals. The module incorporates equipment operating parameters, environmental data, and personnel operation data into a unified collection system.

[0010] S2: A real-time data analysis and processing module, used to process and conduct risk assessment on the data obtained by the multi-source data collection module. The module uses edge computing technology to remove noise data and eliminate outliers at the data collection end through a mean filtering algorithm. For the equipment temperature data sequence {Tt1 , T t2 ,..., T tn}, filtered temperature value A dynamic risk assessment model is constructed using a recurrent neural network, where the hidden layer state h t The update formula is h t = σ(W xh x t + W hh h t-1 + b h ), σ is an activation function such as the tanh function, W xh , W hh are weight matrices, b h is a bias vector, and the output y t The calculation formula is y t = W hy h t + b y , W hy is a weight matrix, b y is a bias vector. The model updates the weight matrix according to real-time data through the backpropagation algorithm to optimize the risk assessment result.

[0011] S3: Risk dynamic adjustment and warning module, used to adjust the risk level and give warnings according to the risk assessment results of the real-time data analysis and processing module. The module sets multiple risk level thresholds {T r1 , T r2 ,..., T rn} based on historical data and industry standards. When the risk assessment result y satisfies T ri ≤ y < T r(i+1) , the risk level is automatically and real-time adjusted to level i.

[0012] S4: Intelligent feedback and optimization module, used to analyze the cause of the accident and optimize the system.

[0013] Preferably, in the multi-source data acquisition module, for key equipment such as reactors, pipelines, and storage tanks, temperature sensors (T s ), pressure sensors (P s ), gas concentration sensors (C s ), and vibration sensors (V s ) are deployed to accurately measure the equipment temperature T, pressure P, specific gas concentration C, and vibration amplitude V.

[0014] Preferably, the accuracy of the temperature sensor can reach ±0.1 °C, the accuracy of the pressure sensor reaches ±0.5% FS, the detection lower limit of the gas concentration sensor is up to the ppm level, the vibration sensor and the displacement sensor have a measurement accuracy of sub-millimeter level, and the motion capture device adopts an optical motion capture system based on computer vision, and captures the operator from different angles through multiple cameras.

[0015] Preferably, the environmental data collection covers the temperature T measured by the temperature and humidity sensor e and the humidity H, the content of harmful gases A measured by the air quality sensor, the light intensity L measured by the light intensity sensor, the noise decibel N measured by the noise level sensor, and the ground settlement displacement D and soil displacement S measured by the displacement sensor. The personnel operation data collection uses the motion capture device to convert the operator's motion trajectory into a motion vector and obtains the operation step sequence O through the docking of the operation process monitoring device and the production system s and the operation time t.

[0016] Preferably, the real-time data analysis and processing module performs deep feature extraction on the preprocessed data. In addition to the equipment temperature change rate it also extracts data statistical features, including but not limited to mean, variance, peak value, frequency domain features, as well as the correlation of environmental parameters and the characteristics of personnel operation behaviors.

[0017] Preferably, the warning method integrates audible and visual alarms, a short message platform and APP push notifications. The notification content includes but not limited to risk type, location, and level information. Different risk levels in different regions are identified by different colors, and the risk distribution and dynamic changes are visually presented using dynamic charts.

[0018] Preferably, the intelligent feedback and optimization module establishes an accident cause analysis model. When a risk warning or a safety accident occurs, the equipment operation parameters, environmental data, and relevant data of personnel operations at the moment of the accident are input. Using the association rule algorithm, by calculating the support and confidence X and Y are parameter subsets, and D is the set of all data. Analyze the correlation between equipment operation parameters and historical fault data to screen the accident causes.

[0019] Preferably, the feedback information is transmitted to the production management department and the operators according to the cause analysis results. The production management department adjusts the production plan and optimizes the process flow accordingly, and the operators improve the operation methods. At the same time, by deeply mining historical data and feedback information, the sensor layout and acquisition frequency of the multi-source data acquisition module are optimized, and the deep learning model parameters of the real-time data analysis and processing module are adjusted using the stochastic gradient descent algorithm.

[0020] Beneficial effects

[0021] Compared with the prior art, the present invention provides a control system for the safe production of hazardous chemicals, which has the following beneficial effects:

[0022] 1. In terms of data collection, the control system for the safe production of hazardous chemicals realizes the all-round collection of equipment operation, environment and personnel operation data. High-precision sensors ensure the accuracy of data, far exceeding the traditional limited-dimensional and precision collection. When analyzing and processing data, edge computing combines with deep learning to dynamically adjust the risk assessment model, breaking through the limitations of traditional fixed models and being able to adapt to production changes in real time. In terms of risk warning, the risk level is dynamically adjusted, with multiple warnings and visual displays. Compared with single warnings and simple presentations, the risk perception is more timely and accurate.

[0023] 2. The control system for the safe production of hazardous chemicals has an intelligent feedback and optimization module. Through a unique accident analysis model, it can find the accurate reasons, promote the improvement of production management and operations, and can also optimize each module of the system, forming a closed-loop improvement mechanism. Overall, the system comprehensively improves the control level of the safe production of hazardous chemicals from data collection to analysis, warning and then to optimization, greatly reducing safety risks and ensuring production safety and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] Please refer to Figure 1 , the present invention proposes a control system for the safe production of hazardous chemicals, including the following:

[0027] S1: Multi-source data collection module

[0028] This module plays a fundamental and crucial role in the control system for the safe production of hazardous chemicals, responsible for comprehensively, highly accurately and in real time collecting various data in the production process of hazardous chemicals. It covers equipment operation parameters, environmental data and personnel operation data. These data constitute the core basis for subsequent risk assessment, analysis and decision-making, ensuring a comprehensive and dynamic perception of the entire production status without dead ends and providing a solid data foundation for safe production control.

[0029] Implementation steps

[0030] S1.1: Equipment operation parameter acquisition: On key equipment in hazardous chemical production workshops, such as reaction kettles, pipelines, storage tanks, etc., install temperature sensors (T s ), pressure sensors (P s ), gas concentration sensors (C s ), vibration sensors (V s ), etc. The temperature sensor (T s ) adopts a high-precision thermistor sensor, whose measurement accuracy can reach ±0.1°C, and can accurately measure the equipment temperature T, providing reliable data for temperature control during the reaction process and monitoring the risk of equipment overheating. The pressure sensor (P s ) selects a capacitive pressure sensor, and the measurement range can be flexibly adjusted according to the actual working pressure of the equipment, with an accuracy of ±0.5% FS, which is used to accurately measure the internal pressure P of the equipment to prevent serious accidents such as explosions caused by abnormal pressure. The gas concentration sensor (C s ) adopts sensors with multiple principles such as catalytic combustion type and electrochemistry for common flammable, explosive, toxic and harmful gases in hazardous chemical production, such as hydrogen, chlorine, hydrogen sulfide, etc., and can accurately measure the specific gas concentration C, and the detection lower limit can reach the ppm level. The vibration sensor (V s ) adopts a piezoelectric vibration sensor, which can effectively measure the vibration amplitude V of the equipment. By analyzing the vibration spectrum, it can detect abnormal vibrations of the equipment caused by mechanical failures, component looseness, etc. in advance and prevent sudden equipment failures.

[0031] S1.2: Environmental data acquisition: Uniformly arrange environmental monitoring sensors in the production area, including temperature and humidity sensors (measuring temperature T e and humidity H), air quality sensors (measuring the content of harmful gases A), light intensity sensors (measuring light intensity L), and noise level sensors (measuring noise decibels N). The temperature and humidity sensor adopts a digital temperature and humidity sensor, which can measure the environmental temperature T eAnd relative humidity H, with accuracies of ±0.5°C and ±3%RH respectively, provides data support for the temperature and humidity control in the storage and production environments of hazardous chemicals, avoiding the impact of abnormal temperature and humidity on product quality or potential safety hazards. The air quality sensor comprehensively applies various detection technologies, such as semiconductor gas sensors, infrared absorption sensors, etc., and can measure the content A of harmful gases in the air in real time, including volatile organic compounds (VOCs), nitrogen oxides, sulfur dioxide, etc., ensuring that the air quality in the production area meets safety standards. The light intensity sensor uses the principle of a photodiode to accurately measure the light intensity L. For some production processes or storage areas of hazardous chemicals that are sensitive to light, the protective measures can be adjusted in a timely manner by monitoring the light intensity. The noise level sensor adopts the principle of a capacitive microphone to measure the noise decibels N, which is used to monitor the noise in the production environment, preventing damage to the hearing of operators caused by excessive noise and operation errors caused by noise interference. At the same time, geological monitoring equipment, such as displacement sensors (measuring the ground settlement displacement D and soil displacement S), is installed in sensitive areas of the surrounding geological conditions. The high-precision fiber Bragg grating displacement sensor can accurately measure the ground settlement displacement D and soil displacement S, with an accuracy up to the sub-millimeter level, providing early warnings for preventing accidents such as the leakage of hazardous chemicals caused by geological disasters.

[0032] S1.3: Personnel operation data collection: Set up motion capture devices in the working areas of operators, record the motion trajectories of operators through image recognition technology, and convert them into motion vectors. The motion capture device adopts an optical motion capture system based on computer vision. By using multiple cameras to shoot operators from different angles and applying advanced image recognition algorithms, it can track the movements of various parts of the operator's body in real time and accurately, and convert them into motion vectors containing information such as position, speed, and acceleration. It is used to analyze the operation standardization and fatigue state of operators. The operation process monitoring device is connected to the production system to obtain the operation step sequence O s and operation time t of the operator in real time. By embedding an operation process monitoring software module in the production system, it can be seamlessly connected to various automated control systems and equipment operation interfaces, accurately record the sequence O s of each step of the operator's operation and the time t spent on the operation, and compare it with the standard operation process to timely detect illegal operation behaviors.

[0033] S2: Real-time data analysis and processing module

[0034] This module is the intelligent core of the entire hazardous chemical production safety control system. Its main function is to efficiently screen, finely preprocess, and deeply extract features from the massive and complex raw data obtained by the multi-source data acquisition module, and build a highly adaptive dynamic risk assessment model by means of advanced deep learning algorithms. By continuously learning and analyzing real-time data, it can quickly and accurately update the risk assessment results according to the dynamic changes of production conditions, environmental factors, and personnel operations, providing a scientific and reliable basis for dynamic risk adjustment and early warning, and achieving precise control of the safety production risks of hazardous chemicals.

[0035] Implementation steps

[0036] S2.1: Edge computing preprocessing: Use edge computing devices to perform preliminary processing on the raw data at the data acquisition end. The mean filtering algorithm is used to remove noise data. For the device temperature data sequence {T t1 , T t2 ,..., T tn} collected at a certain moment t, the filtered temperature value T f is:

[0037] where n is the length of the data sequence, and T ti is the i-th temperature data in the sequence. The mean filtering algorithm can effectively smooth the data, remove random noise caused by factors such as sensor measurement errors and electromagnetic interference, and improve the stability and reliability of the data. At the same time, by setting reasonable thresholds to remove outliers, for example, for temperature data, if T ti exceeds the normal operating temperature range [T min , T max , it is determined as an outlier and excluded to ensure the accuracy of the subsequent analysis data. The edge computing device also has data caching and preliminary aggregation functions. It can temporarily store data when the network transmission is interrupted or congested. After the network returns to normal, the cached data is aggregated according to certain rules and then transmitted to the backend data analysis platform, effectively reducing the data transmission volume and transmission delay, and improving the real-time response ability of the system.

[0038] S2.2: Feature extraction: Extract features from the preprocessed data. For example, for device operating parameters, extract the parameter change rate as a feature. Taking the device temperature change rate R T as an example:

[0039] where T t+1 is the temperature at the next moment, and T tLet \(T\) be the temperature at the current moment and \(\Delta t\) be the time interval. By calculating the temperature change rate, it is possible to detect the abnormal upward or downward trend of the equipment temperature in a timely manner and give early warnings of potential risks such as equipment failures or out-of-control reactions. In addition to the change rate feature, statistical features such as the mean, variance, and peak value of the data can also be extracted, as well as features such as the power spectral density based on frequency domain analysis. For environmental data, features such as the change trend of environmental parameters and the correlation between different parameters can be extracted. For personnel operation data, features such as the speed and acceleration of the action vector and the execution time interval of operation steps can be extracted. These features can comprehensively reflect the operation behavior patterns and states of operators and provide rich information dimensions for risk assessment.

[0040] S2.3: Construction of a deep learning risk assessment model: A dynamic risk assessment model is constructed using a recurrent neural network (RNN). Let the input data sequence be The hidden layer state sequence is The output risk assessment result is

[0041] At time \(t\), the update formula for the hidden layer state \(h\) t is: \(h\) t \(=\sigma(W\) xh \(x\) t \(+W\) hh \(h\) t-1 \(+b\) h )

[0042] where \(\sigma\) is the activation function (such as the tanh function), \(W\) xh is the weight matrix from the input to the hidden layer, \(W\) hh is the weight matrix from the hidden layer to the hidden layer, and \(b\) h is the bias vector. The tanh function can map the input data to the interval \([-1, 1]\), effectively solving the vanishing gradient problem and enabling the model to better learn long-term dependencies. The weight matrices \(W\) xh and \(W\) hh as well as the bias vector \(b\) h are obtained through training and learning with a large amount of historical data. The training process uses the stochastic gradient descent algorithm (SGD) and its variant algorithms, such as Adagrad, Adadelta, RMSProp, Adam, etc., to optimize the model parameters and improve the generalization ability and prediction accuracy of the model.

[0043] The calculation formula for the output \(y\) t is: \(y\) t \(=W\) hy \(h\) t \(+b\) y

[0044] where \(W\) hy is the weight matrix from the hidden layer to the output layer, and \(b\)y is the output layer bias vector. The model continuously inputs real-time data and uses the backpropagation algorithm to update the weight matrix to optimize the risk assessment results. The backpropagation algorithm calculates the error between the predicted result and the true label, and propagates the error from the output layer back to the input layer, thereby adjusting the weight matrix of each layer, so that the model continuously reduces the error during training and improves the accuracy of risk assessment. To prevent the model from overfitting, regularization techniques such as L1 and L2 regularization, as well as Dropout technology, can also be used to randomly discard some neurons and enhance the generalization ability of the model.

[0045] S3: Risk Dynamic Adjustment and Early Warning Module

[0046] Based on the accurate risk assessment results output by the real-time data analysis and processing module, this module quickly and dynamically adjusts the risk level, and issues early warnings to relevant personnel in a diverse and efficient manner in a timely manner to ensure that once a safety risk occurs during the production process of hazardous chemicals, key personnel can be notified within the shortest time. At the same time, with the help of advanced visualization technology, it intuitively and understandably displays the risk distribution and change trends, providing managers with a clear and comprehensive view of the safety status, helping them quickly make scientific decisions, take effective risk response measures, and ensure the safe and stable operation of the hazardous chemical production process.

[0047] Implementation Steps

[0048] S3.1: Risk Level Adjustment: According to historical data and industry standards, set multiple risk level thresholds {T r1 , T r2 ,..., T rn}. The risk level is divided by comprehensively considering various factors such as the nature of hazardous chemicals, the complexity of the production process, the reliability of equipment, the sensitivity of environmental conditions, and the proficiency of personnel operations. For example, for the production process of highly hazardous chemicals, the risk level is divided more meticulously and the threshold is set more strictly. When the risk assessment result y satisfies T ri ≤y<T r(i+1) , the risk level is adjusted to level i. The system will continuously monitor the change of the risk assessment result. Once the risk assessment value crosses the threshold boundary, the risk level will be automatically updated immediately to ensure that the risk level is always consistent with the actual risk situation.

[0049] S3.2: Warning Method: When the risk level reaches the set threshold, activate the audible and visual alarm device. The audible and visual alarm device uses high-brightness flashing lights and high-decibel alarm sounds, which can quickly attract the attention of operators in a noisy production site environment. At the same time, send push notifications to relevant management personnel through the SMS platform or APP. The notification content includes detailed information such as the risk type (such as equipment failure risk, environmental anomaly risk, personnel operation violation risk), risk location (equipment number, area coordinates), and risk level. The SMS platform has high reliability and high arrival rate, ensuring that the notification is delivered to the management personnel's mobile phones in a timely manner. The APP adopts a simple and clear interface design. Management personnel can view the risk details in real time on their mobile phones and can communicate with on-site operators through the APP for instant communication to guide emergency handling work. In addition, the system can also be linked with the enterprise's internal broadcast system, emergency command center, etc. to achieve all-round and multi-level warning information release.

[0050] S3.3: Visualization Display: On the operation interface of the control system, identify the risk levels of different areas with different colors (for example, green represents low risk, yellow represents medium risk, and red represents high risk). Use dynamic charts (such as line charts to show the change trend of risk levels over time, and heat maps to show the distribution of risks in the production area) to intuitively present the risk distribution and dynamic changes. The line chart can clearly show the fluctuations of the risk level over a period of time, helping management personnel analyze the development trend of risks and predict potential risks. The heat map intuitively shows the distribution of risks in the production area through the depth of different colors. The red area represents the area with higher risks, and the blue area represents the area with lower risks. Management personnel can clearly understand the risk situation of the entire production area at a glance, facilitating the targeted deployment of safety prevention measures. In addition, three-dimensional visualization technology can be used to present a three-dimensional display of production equipment, process flows, and risk distributions, presenting the safety status of the production site more realistically and intuitively.

[0051] S4: Intelligent Feedback and Optimization Module

[0052] This module is a key link for the hazardous chemical production safety control system to achieve continuous optimization and continuously improve the safety management level. It can automatically and deeply analyze the reasons according to the risk warnings and actual safety accident situations by using advanced data analysis technology, and transmit accurate feedback information to the production management department and operators in a timely manner. At the same time, by deeply mining historical data and feedback information, optimize the performance of each module of the system from multiple dimensions to form a self-improving and self-enhancing closed-loop optimization mechanism, ensuring that the system is always in an efficient and accurate operating state, and maximizing the safety of hazardous chemical production.

[0053] Implementation Steps

[0054] S4.1: Cause analysis: Establish an accident cause analysis model. When a risk warning or safety accident occurs, relevant data (such as equipment operation parameters, environmental data, and personnel operation data at the time of the accident) are input into the model. For example, for risks caused by equipment failures, by analyzing the correlation between equipment operation parameters and historical failure data, the association rule algorithm is used to find possible causes of the failure. The set of equipment operation parameters is P = {p1, p2,..., p m}, and the set of historical failure data is F = {f1, f2,..., f n}. The association rules are determined by calculating the support support(X→Y) and confidence confidence(X→Y):

[0055]

[0056] where X and Y are subsets of parameters, D is the set of all data, and |·| represents the number of elements in the set. The support reflects the frequency of the rule appearing in the dataset, and the confidence indicates the probability that the conclusion Y appears when the premise condition X is satisfied. The causes related to the accident are screened according to the support and confidence thresholds. In addition to the association rule algorithm, data mining algorithms such as decision trees and Bayesian networks can also be used for cause analysis. By learning a large amount of historical accident data, an accident cause analysis model is constructed to improve the accuracy and efficiency of cause analysis.

[0057] S4.2: Feedback information transmission: Send the analyzed causes to the production management department and relevant operators in the form of a report. The report content includes the time, location, detailed process of the accident, the analyzed causes, and corresponding improvement suggestions. The production management department adjusts the production plan and optimizes the process flow according to the feedback information. For example, if it is found that a certain production link frequently has risks caused by equipment failures, it can consider increasing the equipment maintenance frequency, replacing more reliable equipment, or optimizing the operation process of this link. Operators improve their operation methods according to the feedback, such as learning the correct operation steps through training to improve operation proficiency and standardization. At the same time, the system can also provide targeted training and guidance for operators through online training platforms, operation guide push, etc., to help them quickly master the improved operation methods.

[0058] S4.3: System Optimization: Regularly conduct in-depth mining of historical data and feedback information. For example, optimize the layout and acquisition frequency of sensors in the multi-source data acquisition module according to the occurrence frequency and severity of risks in different regions. For high-risk areas, increase the number and types of sensors and raise the acquisition frequency to obtain more comprehensive and accurate data. For the real-time data analysis and processing module, adjust the parameters of the deep learning model using optimization algorithms (such as the stochastic gradient descent algorithm) based on the accuracy of the model evaluation results to improve the overall system performance and risk assessment accuracy. At the same time, the hardware devices of the system can also be upgraded, such as increasing computing resources and optimizing the network architecture, to adapt to the growing data processing requirements. In addition, by sharing data and exchanging experiences with other enterprises in the same industry, continuously introduce advanced management concepts and technical methods to continuously optimize the system functions and performance.

[0059] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A hazardous chemicals production safety management and control system, characterized by: Includes the following: S1: Multi-source data acquisition module, used to collect data from the production process of hazardous chemicals in an all-round way. The module integrates equipment operating parameters, environmental data and personnel operation data into a unified acquisition system; S2: Real-time data analysis and processing module, used to process and assess the risk of data acquired by the multi-source data acquisition module. The module uses edge computing technology to remove noise data and eliminate abnormal values ​​through a mean filtering algorithm at the data acquisition end. For the equipment temperature data sequence {T t1 ,T t2 ,...,T tn }, filtered temperature value A recurrent neural network is used to construct a dynamic risk assessment model, where the hidden layer state h t Update formula is h t =σ(W xh x t +W hh h t-1 +b h ), σ is an activation function such as tanh function, W xh , W hh is the weight matrix, b h is the bias vector, output y t The calculation formula is y t =W hy h t +b y , W hy is the weight matrix, b y The model uses the back propagation algorithm to update the weight matrix based on real-time data to optimize the risk assessment results. S3: Risk dynamic adjustment and early warning module, used to adjust the risk level and issue early warning according to the risk assessment results of the real-time data analysis and processing module. The module sets multiple risk level thresholds based on historical data and industry standards. r1 ,T r2 ,...,T rn }, when the risk assessment result y satisfies T ri ≤y<T r(i+1) When the risk level is adjusted to level i automatically and in real time; S4: Intelligent feedback and optimization module, used to analyze the cause of the accident and optimize the system.

2. A hazardous chemicals production safety management and control system according to claim 1, characterized in that: In the multi-source data acquisition module, temperature sensors (T s )、Pressure sensor (P s )、Gas concentration sensor (C s )、Vibration sensor (V s ) to accurately measure the device temperature T, pressure P, specific gas concentration C, and vibration amplitude V.

3. A hazardous chemicals production safety management and control system according to claim 2, characterized in that: The temperature sensor has an accuracy of ±0.1°C, the pressure sensor has an accuracy of ±0.5% FS, the gas concentration sensor has a detection limit of ppm, the vibration sensor and displacement sensor have sub-millimeter measurement accuracy, and the motion capture device uses an optical motion capture system based on computer vision to shoot operators from different angles through multiple cameras.

4. A hazardous chemicals production safety management and control system according to claim 3, characterized in that: The environmental data collection includes the temperature T measured by the temperature and humidity sensor e and humidity H, harmful gas content A measured by air quality sensor, light intensity L measured by light intensity sensor, noise decibel N measured by noise level sensor, and ground subsidence displacement D and soil displacement S measured by displacement sensor. Personnel operation data collection converts operator motion trajectory into motion vector with the help of motion capture equipment. And obtain the sequence of operation steps through the operation process monitoring device and the production system. s and operation time t.

5. A hazardous chemicals production safety management and control system according to claim 1, characterized in that: The real-time data analysis and processing module performs deep feature extraction on the pre-processed data, except for the device temperature change rate. The data statistical features are also extracted, including but not limited to mean, variance, peak value, frequency domain features, environmental parameter correlation, and personnel operation behavior characteristics.

6. A hazardous chemicals production safety management and control system according to claim 1, characterized in that: The early warning method integrates sound and light alarms, SMS platforms and APP push notifications. The notification content includes but is not limited to risk type, location, and level information. Different colors are used to identify the risk levels of different areas, and dynamic charts are used to intuitively present the risk distribution and dynamic changes.

7. A hazardous chemicals production safety management and control system according to claim 1, characterized in that: The intelligent feedback and optimization module establishes an accident cause analysis model. When a risk warning or safety accident occurs, the equipment operating parameters, environmental data and relevant data of personnel operations at the time of the accident are input, and the association rule algorithm is used to calculate the support degree. and confidence X and Y are parameter subsets, and D is the set of all data. The correlation between equipment operating parameters and historical fault data is analyzed to screen the causes of accidents.

8. A hazardous chemicals production safety management and control system according to claim 7, characterized in that: The feedback information is transmitted to the production management department and operators based on the cause analysis results. The production management department adjusts the production plan and optimizes the process flow accordingly, and the operators improve their operating methods. At the same time, through deep mining of historical data and feedback information, the sensor layout and acquisition frequency of the multi-source data acquisition module are optimized, and the stochastic gradient descent algorithm is used to adjust the deep learning model parameters of the real-time data analysis and processing module.

Images