Enterprise production safety monitoring and checking system and method based on multi-source data fusion

Abstract

The invention provides an enterprise production safety monitoring and checking system based on multi-source data fusion, and the system is characterized in that the system comprises a data collection module which collects production environment data, processes the data, and generates a time-space alignment data set; the feature screening module is used for screening features highly related to security from the space-time alignment data set to generate a feature matrix; the model construction module is used for constructing a risk assessment model, calculating a dynamic risk value according to the multi-dimensional features, and performing risk grading on the dynamic risk value; the response control module is used for starting a corresponding response strategy according to the risk classification and outputting an early warning instruction set and an equipment control signal; the monitoring calculation module is used for monitoring control, comparing risk value changes before and after treatment and calculating response efficiency; and the adjusting and optimizing module is used for carrying out dynamic adjustment according to the response efficiency and optimizing the response strategy. And the limitation of traditional single-dimensional monitoring is broken through, accurate correlation analysis of equipment, environment and personnel risks is realized, and the composite hidden danger recognition capability is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of enterprise production safety, in particular to an enterprise production safety monitoring and investigation system and method based on multi-source data fusion. BACKGROUND

[0002] Enterprise production safety refers to a comprehensive system that prevents, controls and eliminates risks that may cause personal injury, equipment damage or environmental pollution in production activities through systematic management methods and technical measures, and ensures the smooth operation of the production process. Its core lies in establishing a four-dimensional integrated prevention and control mechanism covering people, machines, environment and management: improving employees' risk awareness and disposal ability at the personnel level; monitoring the running state of equipment in real time to prevent mechanical failure at the equipment level; dynamically monitoring the temperature, gas concentration and dust hazards in the work area at the environmental level; and achieving closed-loop management through risk assessment and hidden danger investigation at the management level. The fundamental goal is to reduce the accident rate, protect the safety of workers, maintain the integrity of enterprise assets and social sustainable development, and reflect the social responsibility and governance level of the enterprise.

[0003] The traditional enterprise production safety monitoring scheme usually adopts a single-dimensional, static and artificial-dependent management mode, which has significant limitations: data acquisition relies on a single sensor or manual inspection to obtain equipment temperature, pressure basic parameters, lacks multi-source data cooperation of video monitoring, personnel positioning and environmental indicators, resulting in one-sided risk perception; the risk assessment adopts a fixed threshold alarm mechanism, which cannot dynamically adjust the risk weight in combination with real-time working conditions, and it is difficult to identify complex hidden danger response strategies; after the early warning, it relies on manual judgment of disposal measures, the response delay is long and is easily disturbed by subjective factors, and it is difficult to achieve hierarchical control; closed-loop management is missing, there is a lack of disposal effect quantitative evaluation and strategy self-optimization mechanism, and historical data cannot drive model iteration, similar accidents repeatedly occur. These defects lead to high false alarm rate, response lag and poor adaptability of the traditional scheme, especially in the face of high-risk operations and complex scenes of sudden accidents, it is difficult to achieve precise prevention and control and rapid loss stop. SUMMARY

[0004] To solve the above technical problems, the technical scheme adopted by the present application is: an enterprise production safety monitoring and investigation system based on multi-source data fusion, characterized by comprising:

[0005] A data acquisition module acquires production environment data, processes the data, and generates a spatio-temporal alignment data set;

[0006] A feature selection module selects features strongly related to safety from the spatio-temporal alignment data set and generates a feature matrix;

[0007] A model construction module constructs a risk assessment model, calculates a dynamic risk value according to multi-dimensional features, and classifies the dynamic risk value by risk level;

[0008] The response control module starts a corresponding response strategy according to the risk classification, and outputs a warning instruction set and a device control signal;

[0009] The monitoring calculation module monitors the control, compares the risk value changes before and after the treatment, and calculates the response efficiency;

[0010] The adjustment optimization module dynamically adjusts according to the response efficiency, and optimizes the response strategy.

[0011] The collected production environment data includes: environmental parameters including temperature and humidity, air pressure noise, light intensity, gas concentration, dust particles; spatial positioning data including personnel activity area, material flow path, equipment vibration and displacement; device parameters including current and voltage fluctuation, internal temperature and humidity of key equipment, and state of cooling fan.

[0012] The generated multi-dimensional feature matrix is specifically: a structured matrix including three dimensions of device state, environmental index and personnel behavior is constructed; the key features extracted in the device state dimension include device running time, downtime frequency, abnormal vibration frequency and energy consumption deviation; the key features extracted in the environmental index dimension include average gas concentration in dangerous area, temperature and humidity fluctuation threshold and dust accumulation rate; the key features extracted in the personnel behavior dimension include high-risk area stay time, violation operation frequency and emergency exercise response speed; the screening standard is to select features strongly correlated with historical accidents, and only numerical value indexes are retained.

[0013] The way to calculate the dynamic risk value is: combining real-time data and historical trends, the expression can be used:

[0014]

[0015] Where R is the dynamic risk value, wi is the preset feature weight, Si is the device environment monitoring value deviation, Pj is the personnel behavior event intensity, Tj is the behavior time attenuation factor, and a and b are the dynamic weights of device environment and personnel behavior.

[0016] The setting method of the dynamic weights of device environment and personnel behavior is: different weights are set according to different risk scenarios; the initial set weight is used in daily monitoring scenario; the high-risk operation scenario includes physical high-risk environment, chemical energy risk and special process activity, and the environmental index weight needs to be increased; the emergency response scenario includes natural disaster event, production accident event and network security event, and the personnel behavior weight needs to be increased.

[0017] The risk grading manner is that when the dynamic risk value R is less than a threshold value x, marked as blue, daily monitoring is performed on production, and data is recorded in real time; when the dynamic risk value R is greater than the threshold value x and less than a threshold value y, marked as yellow, an early warning is pushed to a manager, and the frequency of inspection is increased; when the dynamic risk value R is greater than the threshold value y and less than a threshold value z, marked as orange, the device is operated at a reduced speed, an abnormal area is isolated, and a worker is notified to evacuate quickly; and when the dynamic risk value R is greater than the threshold value z, marked as red, production is immediately stopped, and an emergency system including spraying and air exhaust is started.

[0018] The manner of calculating the response efficiency is that the response efficiency is calculated according to the change of the risk values before and after treatment, which can be calculated by the expression:

[0019]

[0020] wherein E is the response efficiency, R1 is the risk value before treatment, R2 is the risk value after treatment, and T is the response duration.

[0021] The manner of optimizing the model is that when the response efficiency is greater than a preset threshold value V, the current response process is maintained and is popularized to similar scenarios; when the response efficiency is less than the preset threshold value V and greater than a preset threshold value U, the personnel are trained and the equipment maintenance cycle is shortened; and when the response efficiency is less than the preset threshold value U, the emergency plan needs to be redesigned, and the response time of a key link is shortened.

[0022] The enterprise production safety monitoring and troubleshooting method based on multi-source data fusion comprises a data acquisition end, a feature screening end, a model construction end, a response control end, a monitoring calculation end and an adjustment optimization end, and specifically comprises the following steps:

[0023] S1, collecting production environment data, processing the data, and generating a spatio-temporal alignment data set;

[0024] S2, screening safety strongly related features from the spatio-temporal alignment data set, and generating a multi-dimensional feature matrix;

[0025] S3, constructing a risk assessment model, calculating a dynamic risk value according to the multi-dimensional features, and grading the dynamic risk value;

[0026] S4, starting a corresponding response strategy according to the risk grading, and outputting an early warning instruction set and a device control signal;

[0027] S5, monitoring the control, comparing the change of the risk values before and after treatment, and calculating the response efficiency;

[0028] S6, dynamically adjusting according to the response efficiency, and optimizing the response strategy.

[0029] In combination with all the technical solutions described above, the application has the following positive effects: 1. The application uses multi-dimensional data sources of integrated equipment operation, environmental monitoring and personnel behavior, adopts a space-time alignment algorithm to eliminate the space-time deviation of heterogeneous data, constructs a standardized data pool with global coverage, and compared with traditional single-dimensional monitoring, multi-source data fusion can accurately capture the composite risks of equipment abnormalities, environmental mutations and personnel violations, break through the perception limitations of a single data source, realize cross-dimensional risk correlation analysis, and improve the coverage rate of hidden danger identification.

[0030] 2. The application builds a multi-dimensional feature matrix, dynamically adjusts the risk weight in combination with real-time working conditions, establishes a risk assessment model with scene self-adaptation, breaks through the rigidity of traditional fixed threshold alarm, identifies the superposition effect of equipment gradual failure and sudden events, reduces the false alarm rate while improving the early warning sensitivity of high-risk scenes, significantly reduces the false alarm rate, and improves the early identification ability of gradual failure.

[0031] 3. The application automatically triggers a differentiated control strategy through a hierarchical response mechanism, pushes artificial inspection instructions for low-risk, and links equipment regulation for medium and high-risk, replaces the traditional manual confirmation process through an automatic decision-making closed loop, improves the response rate, reduces the response time of high-risk scenes, and effectively suppresses the probability of accidents.

[0032] 4. The application calculates the response efficiency by quantifying the risk value change rate before and after disposal, dynamically adjusts the model parameters and response rules in combination with a reinforcement learning algorithm, forms a closed loop mechanism of collection, analysis, response and feedback, enables the system to update the strategy based on historical disposal effects, continuously improves the risk early warning accuracy and emergency response efficiency, and breaks through the capability solidification bottleneck of traditional static systems. BRIEF DESCRIPTION OF DRAWINGS

[0033] The application will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the application. For ordinary skilled persons in the art, other drawings can be obtained without creative labor on the basis of the following drawings.

[0034] Figure 1 The system framework diagram of the system of the application. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by ordinary skilled persons in the art without creative labor are within the protection scope of the application.

[0036] Reference is made to Figure 1As shown, the present application proposes an enterprise production safety monitoring and troubleshooting system based on multi-source data fusion, including a data acquisition module, a feature screening module, a model construction module, a response control module, a monitoring calculation module, and an adjustment and optimization module.

[0037] In a more specific application of the present application, the collection of enterprise production environment data uses Internet of Things technology to build a global perception network, and realizes real-time monitoring through the deployment of multiple types of sensors and intelligent terminals. For environmental parameters, wireless temperature and humidity sensors, high-precision barometers, noise decibel meters, photoelectric light sensors, electrochemical or infrared gas detectors, and laser scattering particulate matter monitors are used to continuously sample the physical and chemical environment of the work area; spatial positioning data is obtained by fusing UWB ultra-wideband positioning base stations, RFID radio frequency identification tags, and acceleration sensors, which can track the activity trajectory of personnel, the transportation path of materials, and the vibration displacement state of equipment in real time, and combine laser range finders and gyroscopes to calibrate the accuracy of spatial coordinates; device parameter monitoring integrates current transformers, voltage transducers, embedded temperature and humidity probes, and speed sensors to collect signals for current and voltage fluctuations, internal micro-environment, and cooling system operation status of key equipment.

[0038] All data are pre-processed through edge computing by industrial gateways, time synchronization protocol is used to eliminate clock drift, and laser space calibration technology is used to realize three-dimensional coordinate alignment of multi-source data, which is uploaded to the central database through OPCUA protocol to form a unified spatio-temporal monitoring data stream of all factors, providing a high-quality data basis for subsequent analysis.

[0039] From the spatio-temporal alignment data set, features strongly related to safety are selected to generate a multi-dimensional feature matrix. The original data stream can be decoupled according to the device, environment, and personnel dimensions: key parameters reflecting the health status of the device are extracted from the device log, including running time and downtime frequency; indicators representing environmental stability are calculated from the environmental sensor sequence, including gas concentration mean and temperature and humidity fluctuation threshold; and behavior pattern features are calculated from personnel trajectory data, including high-risk area stay time and frequency of illegal operations. In this process, the quantifiability of the features needs to be strictly verified, and all qualitative descriptions that cannot be converted into numerical values are removed.

[0040] Screening core features through relevance analysis of historical accident library: For each accident data slice, the mutual information algorithm is used to quantify the statistical correlation between features and accident labels, such as the mutual information value of the gas concentration rising rate and the valve opening deviation feature in the dangerous chemical leakage accident, which is significantly higher than other parameters, indicating its strong correlation with safety. At the same time, the random forest model is used to train the feature importance ranking, and the redundant items with low contribution to risk prediction are removed, such as the change of light intensity in the conventional area, which is filtered out because it has no significant causal relationship with the accident. In this stage, the spatiotemporal consistency constraint needs to be applied to verify the physical logic relationship between features: when the equipment vibration exceeds the limit event occurs, check whether the personnel are in the high-risk area and the environmental parameters are abnormal at the same time window, and exclude the pseudo-correlation features caused by data collection errors.

[0041] After completing the feature screening, a structured matrix of three-dimensional intersection of equipment state, environmental indicators, and personnel behavior is constructed. The row dimension of the matrix is divided by the physical space unit, and the column dimension is composed of key features of each dimension. The matrix data is dynamically updated in a rolling time window, and is continuously optimized through the decay rule: if a feature has not triggered a warning for a long time, it is automatically downgraded to an observation item; when a new risk pattern appears, the incremental learning mechanism is triggered to add the corresponding feature column, ensuring that the matrix always covers the latest risk pattern. This process realizes the deep adaptation of feature engineering and production safety scene through the closed-loop logic of risk tracing, quantitative screening, spatiotemporal verification, and dynamic iteration.

[0042] The way to calculate the dynamic risk value is: combining real-time data and historical trends, the expression can be:

[0043]

[0044] Where R is the dynamic risk value, wi is the preset feature weight, Si is the equipment environment monitoring value deviation, Pj is the personnel behavior event intensity, Tj is the behavior time decay factor, and a, b are the dynamic weights of equipment environment and personnel behavior.

[0045] The setting method of the dynamic weights of equipment environment and personnel behavior is: according to different risk scenarios, different weights are set; the daily monitoring scene follows the initial set weight; the high-risk operation scene includes physical high-risk environment, chemical energy risk, and special process activity, which needs to increase the weight of environmental indicators; the emergency response scene includes natural disaster events, production accident events, and network security events, which need to increase the weight of personnel behavior.

[0046] The risk grading manner is as follows: when the dynamic risk value R is less than a threshold value x, marked as blue, daily monitoring is performed on production, and data is recorded in real time; when the dynamic risk value R is greater than the threshold value x and less than a threshold value y, marked as yellow, an early warning is pushed to a manager, and the frequency of inspection is increased; when the dynamic risk value R is greater than the threshold value y and less than a threshold value z, marked as orange, the device is operated at a reduced speed, an abnormal area is isolated, and a worker is notified to evacuate quickly; and when the dynamic risk value R is greater than the threshold value z, marked as red, production is immediately stopped, and an emergency system including spraying and air exhaust is started. Through multi-source data fusion and dynamic weight adjustment, the limitations of traditional single-dimensional monitoring are broken through, precise correlation analysis of device, environment and personnel risks is achieved, and the identification capability of complex hidden dangers is significantly improved.

[0047] In a specific embodiment, in a chemical production scene, in a reaction kettle area of a certain chemical plant, an abnormality occurs in a daily production process, the weight of an environmental index is set as 0.6, the weight of personnel behavior is set as 0.3, the weight of a device state is set as 0.1, the dynamic weight is distributed, the device environment is 0.7, and the personnel behavior is 0.3; the features are quantified, the current deviation degree is 3, the gas concentration deviation degree is 2.5, the illegal operation intensity is 2, and the time attenuation factor is 0.78; the dynamic risk value R is calculated as 2.03, the dynamic risk value is normalized to obtain R as 81.2. The risk is graded, the threshold value z is set as 80, the reaction kettle power is immediately cut off when the red early warning is reached, the emergency air exhaust and spraying systems are started, the evacuation instruction is broadcast, and a 50-meter area is blocked. The accident information is pushed to an emergency management department.

[0048] The control is monitored, the risk value changes before and after disposal are compared, the response efficiency is calculated, and the expression is as follows:

[0049]

[0050] Wherein E is the response efficiency, R1 is the risk value before disposal, R2 is the risk value after disposal, and T is the response duration.

[0051] When the response efficiency is greater than a preset threshold value V, the current response process is maintained, and the response process is popularized to similar scenes; when the response efficiency is less than the preset threshold value V and greater than a preset threshold value U, the personnel are trained and the equipment maintenance cycle is shortened; when the response efficiency is less than the preset threshold value U, the emergency plan needs to be redesigned, and the response time of a key link is shortened. Relying on a closed-loop feedback mechanism, the response efficiency quantitative evaluation and strategy iteration are combined, the emergency plan is continuously optimized, the disposal delay of a high-risk scene is effectively shortened, a whole-process self-evolution system of risk perception, intelligent decision and efficiency improvement is formed, and the active prevention and control capability of enterprise production safety is comprehensively improved.

[0052] In a specific embodiment, the pre-treatment risk value is 85, the post-treatment risk value is 25, the response duration is 15 minutes, the response efficiency is 4.0, the response efficiency exceeds the preset threshold, and it is a high-efficiency scene, the spatio-temporal features of the current response strategy are extracted and stored in the standard strategy library, and the applicable scene is labeled; if the response efficiency is a medium-efficiency scene, the treatment delay link is extracted, a customized training module is generated accordingly, and the relevance of equipment failure is analyzed to shorten the maintenance cycle; if the response efficiency is a low-efficiency scene, the bottleneck is located through association rule mining, the key nodes are verified by combining fault tree analysis, the mechanical response link is shortened, and the emergency path is reconstructed.

[0053] The response efficiency is calculated by quantifying the change rate of the risk value before and after treatment, the model parameters and response rules are dynamically adjusted by combining the reinforcement learning algorithm, a closed-loop mechanism of collection, analysis, response and feedback is formed, the system can update the strategy based on the historical treatment effect, continuously improve the risk warning accuracy and emergency response efficiency, and break through the capability solidification bottleneck of the traditional static system.

[0054] The enterprise production safety monitoring and troubleshooting method based on multi-source data fusion comprises a data acquisition end, a feature screening end, a model construction end, a response control end, a monitoring calculation end and an adjustment optimization end, and specifically comprises the following steps:

[0055] S1, collecting production environment data, processing the data, and generating spatio-temporal alignment data set;

[0056] S2, screening safety strongly related features from the spatio-temporal alignment data set, and generating a multi-dimensional feature matrix;

[0057] S3, constructing a risk assessment model, calculating a dynamic risk value according to the multi-dimensional features, and classifying the dynamic risk value;

[0058] S4, starting the corresponding response strategy according to the risk classification, and outputting the early warning instruction set and the equipment control signal;

[0059] S5, monitoring the control, comparing the risk value before and after treatment, and calculating the response efficiency;

[0060] S6, dynamically adjusting according to the response efficiency, and optimizing the response strategy.

[0061] Secondly, in the drawings of the disclosed embodiments, only the structures related to the disclosed embodiments are involved, other structures can be referred to the general design, and the same embodiments and different embodiments of the present application can be combined with each other without conflict;

[0062] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. An enterprise production safety monitoring and troubleshooting system based on multi-source data fusion, characterized in that, The method comprises the following steps: a data acquisition module acquires production environment data, processes the data, and generates a spatio-temporal alignment data set; a feature screening module screens features strongly related to safety from the spatio-temporal alignment data set, and generates a feature matrix; a model construction module constructs a risk assessment model, calculates a dynamic risk value according to multi-dimensional features, and classifies the dynamic risk value; a response control module starts a corresponding response strategy according to the risk classification, and outputs a warning instruction set and a device control signal; a monitoring calculation module monitors the control, compares the risk value changes before and after the treatment, and calculates the response efficiency; an adjustment and optimization module dynamically adjusts and optimizes the response strategy according to the response efficiency.

2. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion according to claim 1, characterized in that: The acquisition of production environment data includes: environmental parameters including temperature and humidity, air pressure, noise, light intensity, gas concentration, dust particles; spatial positioning data including personnel activity area, material flow path, device vibration and displacement; device parameters including current and voltage fluctuation, internal temperature and humidity of key devices, and state of cooling fans.

3. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion according to claim 1, characterized in that: The generation of the multi-dimensional feature matrix specifically comprises: constructing a structured matrix including three dimensions of device state, environmental index, and personnel behavior; the key features extracted in the device state dimension include device running time, shutdown frequency, abnormal vibration frequency, and energy consumption deviation; the key features extracted in the environmental index dimension include average gas concentration in dangerous areas, temperature and humidity fluctuation threshold, and dust accumulation rate; the key features extracted in the personnel behavior dimension include high-risk area stay time, frequency of illegal operation, and emergency response speed; the screening standard is to select features strongly related to historical accidents, and only numerical indicators are retained.

4. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion of claim 1, wherein: The manner of calculating the dynamic risk value is: in combination with real-time data and historical trends, the expression: Wherein R is the dynamic risk value, wi is the preset feature weight, Si is the device environment monitoring value deviation, Pj is the personnel behavior event intensity, Tj is the behavior time attenuation factor, and a and b are the dynamic weights of device environment and personnel behavior.

5. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion according to claim 4, characterized in that: The setting method of the dynamic weights of device environment and personnel behavior is: setting different weights according to different risk scenarios; the daily monitoring scenario follows the initial set weights; the high-risk operation scenario includes physical high-risk environment, chemical energy risk, and special process activity, and the environmental index weight needs to be increased; the emergency response scenario includes natural disaster events, production accident events, and network security events, and the personnel behavior weight needs to be increased.

6. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion according to claim 1, characterized in that: The risk classification method is: when the dynamic risk value R is less than the threshold value x, it is marked as blue, the production is monitored daily, and the data is recorded in real time; when the dynamic risk value R is greater than the threshold value x and less than the threshold value y, it is marked as yellow, the warning is pushed to the management personnel, and the inspection frequency is increased; when the dynamic risk value R is greater than the threshold value y and less than the threshold value z, it is marked as orange, the device runs at a reduced speed, the abnormal area is isolated, and the staff is notified to evacuate quickly; when the dynamic risk value R is greater than the threshold value z, it is marked as red, and the production is immediately stopped, and the emergency system including spraying and exhaust is started.

7. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion of claim 1, wherein: The response efficiency is calculated according to the changes of the risk values before and after the treatment, which can be expressed as: wherein E is the response efficiency, R1 is the risk value before the treatment, R2 is the risk value after the treatment, and T is the response time length.

8. The enterprise production safety monitoring and troubleshooting system based on multi-source data fusion of claim 1, wherein: The optimization manner of the model is: when the response efficiency is greater than a preset threshold V, the current response process is kept and is popularized to similar scenarios; when the response efficiency is less than the preset threshold V and greater than a preset threshold U, personnel are trained and the equipment maintenance period is shortened; and when the response efficiency is less than the preset threshold U, the emergency plan needs to be redesigned, and the response time of a key link is shortened.

9. An enterprise production safety monitoring and troubleshooting method based on multi-source data fusion, characterized in that, The method comprises a data acquisition end, a feature screening end, a model construction end, a response control end, a monitoring calculation end and an adjustment optimization end, and specifically comprises the following steps: S1, collecting production environment data, processing the data, and generating a time-space aligned data set; S2, screening safety strongly related features from the time-space aligned data set, and generating a multi-dimensional feature matrix; S3, constructing a risk assessment model, calculating a dynamic risk value according to the multi-dimensional features, and performing risk grading on the dynamic risk value; S4, starting a corresponding response strategy according to the risk grading, and outputting a warning instruction set and an equipment control signal; S5, monitoring the control, comparing the risk value changes before and after the treatment, and calculating the response efficiency; S6, dynamically adjusting according to the response efficiency, and optimizing the response strategy.

Images