Occupational health risk multi-dimensional dynamic evaluation and intelligent management system based on industrial internet

Abstract

The invention discloses an occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the industrial internet, and the system comprises a multi-dimensional data collection module which collects environment parameters, industrial equipment operation parameters and the individual data of operators in real time through industrial internet equipment, a positioning system is combined to calculate individual real-time exposure of operators, and a multi-source data association mechanism of the environment, the equipment and the operators is constructed; the dynamic evaluation model module constructs a dynamic evaluation model based on multi-source data, and dynamically optimizes an individual risk baseline through a federated learning framework to solve the hysteresis of traditional static threshold evaluation, in the invention, the data dimension is more comprehensive, through environment, equipment and individual multi-source data association, the limitation of traditional single environment parameters is broken through, and the evaluation efficiency is improved. The risk assessment accuracy is improved; the evaluation model is dynamic, individual risk baseline self-adaption is realized based on federated learning, a risk source is accurately positioned in combination with causal inference, and the problem of static threshold hysteresis is solved.

Description

Technical Field

[0001] This invention belongs to the field of industrial internet technology, and in particular relates to a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the industrial internet. Background Technology

[0002] Existing occupational health management systems primarily rely on environmental sensors and health check data to build a basic monitoring system. At the environmental monitoring level, these systems typically collect physicochemical parameters using equipment such as temperature, humidity, dust concentration, and hazardous gas detectors. These parameters are then combined with indicators from regular employee health checkups, such as blood tests and lung function tests, to form basic risk assessment data. Technically, they generally employ centralized databases to store data and rely on manually set fixed thresholds for risk warnings, such as setting equipment vibration thresholds based on ISO 10816-1 standards or classifying noise exposure levels according to GBZ 2.1-2019. While these systems have achieved real-time monitoring of basic environmental parameters in industries such as chemical engineering and manufacturing, and have implemented some work order management functions through integration with ERP systems, they have not yet deeply integrated equipment operation data (such as machine tool spindle torque and production line cycle time) and production process information. The data dimension mainly remains at the environmental-physiological binary structure, lacking the ability to dynamically model human-machine collaboration scenarios.

[0003] As Industry 4.0 advances, the limitations of traditional systems are becoming increasingly apparent. First, a single environmental parameter cannot reflect the comprehensive risks under complex operating conditions. For example, the combined effects of arc radiation, electromagnetic interference, and metal dust in a welding workshop are difficult to capture with a single sensor, resulting in a false alarm rate as high as 32%. Second, fixed threshold models cannot adapt to individual differences. For instance, the varying tolerances of employees in different positions within the new energy industry to organic solvents cannot be identified using a unified standard, leading to a false alarm rate of 28%. Regarding intervention mechanisms, existing systems mostly adopt an "alarm-manual response" model, with response times in high-risk scenarios generally exceeding 5 minutes, failing to meet the millisecond-level equipment control requirements of smart factories. In terms of data security, centralized storage models result in a 92% risk of raw data leakage, and cross-departmental data sharing lacks privacy protection mechanisms, making it difficult to meet compliance requirements such as the Personal Information Protection Law. Furthermore, the system lacks compatibility with industrial protocols (such as OPCUA and Profinet), supporting only communication between 1-2 types of devices, making it difficult to achieve health-efficiency collaborative optimization in new production scenarios such as digital twins and human-machine collaboration. Summary of the Invention

[0004] To overcome the aforementioned shortcomings of existing technologies, embodiments of the present invention provide a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, which solves the problems of single data, static assessment, slow response, poor security, and insufficient compatibility in traditional technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet includes:

[0007] The multi-dimensional data acquisition module collects environmental parameters (dust concentration, harmful gas concentration, noise decibel value, vibration value), individual physiological data (heart rate variability, skin conductance response, electromyography signal), and equipment operating parameters (torque, current fluctuation, vibration frequency) in real time through environmental sensors, wearable physiological monitoring devices, industrial equipment operation monitoring modules, and positioning base stations. It uses the IEEE1588PTP protocol to achieve time synchronization and builds a multi-source data association mechanism between the environment, equipment, and people.

[0008] The dynamic assessment model module constructs a dynamic assessment model based on multi-source data. It aggregates encrypted gradient parameters from multiple factories through a horizontal federated learning architecture, optimizes individual risk baselines (distinguishing between age and medical history differences), identifies risk causal relationships using a causal graph model (structural equation model), and achieves advance prediction of risk trends in the next 2-4 hours through a time-series prediction model (LSTM neural network).

[0009] The intelligent management closed-loop module links the industrial production control system (MES) and the resource scheduling system (AGV). Through the edge computing gateway (supporting OPCUA, Profinet, and ModbusTCP protocol conversion), it executes hierarchical intervention measures (level 1 early warning, level 2 equipment control, and level 3 production suspension). It verifies the intervention effect in real time and feeds it back to the evaluation model for iterative optimization. At the same time, it integrates digital twin sand table simulation and human-machine collaborative risk prevention and control functions.

[0010] The cross-domain collaboration module uses privacy computing technologies (homomorphic encryption, differential privacy) to achieve secure data sharing among multiple entities. It uses blockchain to store key operation logs and builds an occupational health management collaboration system covering enterprises, industries, and medical and social security systems. It also supports integration of third-party systems through RESTful APIs.

[0011] Preferably, the multi-dimensional data acquisition module includes:

[0012] The individual exposure quantification unit, based on the distribution of environmental sensors in a 5m×5m grid in the workshop and UWB positioning trajectory, calculates the cumulative exposure of individuals through a path integral algorithm (cumulative exposure = Σ area concentration × residence time), and dynamically adjusts the exposure threshold by associating with individual physiological characteristic parameters (tolerance to organic solvents, sensitivity to electromagnetic radiation) (the threshold is reduced by 30% for individuals with a history of asthma).

[0013] The multimodal data fusion unit integrates physiological data collected by wearable devices, behavioral data (operation posture, protective equipment wearing status) collected by industrial vision devices (based on YOLOv8 algorithm), and environmental sensor data to form a three-dimensional data matrix of "physiological state-behavioral norms-environmental exposure intensity", with a data update frequency of 1 time / second.

[0014] Preferably, the dynamic evaluation model module includes:

[0015] The Federated Learning Baseline Optimization Unit uses the FedProx algorithm to train models locally at the edge nodes of each plant area, and only uploads encrypted gradient parameters (Paillier homomorphic encryption). After aggregation, it generates differentiated risk baselines (noise tolerance baseline of 85dB for young workers and 75dB for elderly workers).

[0016] The causal tracing unit analyzes the correlation between multi-source data using the Do-Calculus algorithm, distinguishes between the correlation and causality of risk factors, accurately locates the core risk sources (environmental factors, equipment factors, or human factors), and outputs a visualized result of the risk causal chain.

[0017] Preferably, the intelligent management closed-loop module includes:

[0018] The tiered intervention implementation unit executes three levels of intervention based on risk level:

[0019] Level 1: Vibration and voice alerts are pushed through wearable devices (response time < 2 seconds).

[0020] Level 2: Adjust the operating parameters of the linkage equipment (reduce welding current by 10%, increase ventilation system frequency to 50Hz);

[0021] Level 3: Trigger the production system to pause and lock the equipment, and push AR evacuation routes (response time <30 seconds).

[0022] The effect verification iteration unit monitors changes in environmental parameters, equipment status, and physiological data in real time after intervention, calculates the intervention effect index (EEI), and triggers model parameter updates when EEI < 0.7 (iteration cycle < 2 hours).

[0023] Preferably, the cross-domain collaboration module includes:

[0024] The privacy-preserving data sharing unit uses secure multi-party computation (MPC) to achieve cross-plant data collaborative analysis, adds Laplace noise to individual health data (meeting the ε=0.5 differential privacy standard), and only shares model parameters and analysis results;

[0025] The social service linkage unit connects with the hospital's PACS system via the FHIR protocol and links with the social security system via the API interface. When an individual's cumulative exposure exceeds the standard, it automatically triggers an occupational disease examination appointment and simultaneously adjusts the work injury insurance rate (20% increase for high-risk groups).

[0026] Preferably, the intelligent management closed-loop module also includes a digital twin sand table simulation unit, which builds a virtual factory image with an accuracy of 0.1mm based on Unreal Engine, maps 80% of the equipment status in real time, simulates the fatigue index (NASA-TLX scale) and cumulative exposure under different shift schedules (three shifts / two shifts), and shortens the emergency response time to less than 90 seconds by overlaying virtual evacuation routes through HoloLens2.

[0027] Preferably, the intelligent management closed-loop module also includes a human-machine collaboration risk prevention and control unit, which integrates the motion trajectory data (TCP coordinates, speed) of the industrial robot (KUKAKR10R1420) with the UWB positioning data of the operator (accuracy ±0.1m), and predicts the collision probability in the next 3 seconds through an LSTM model. When the probability is >0.5, the robot joint impedance is reduced by 50%, and when it is >0.7, an emergency stop is triggered and an avoidance prompt is pushed (response time <200ms).

[0028] Preferably, the path integral algorithm of the individual exposure quantification unit performs numerical integration using the Gaussian quadrature method (time step 1 second) to calculate the sum of the products of the worker's stay time in each area and the real-time environmental parameter concentration of the corresponding area. The environmental parameters include the concentration of benzene series compounds collected by the multispectral gas detector and the vibration value collected by the IEPE accelerometer.

[0029] Preferably, the industrial internet device includes:

[0030] Environmental sensors (SensirionSHT45 temperature and humidity sensor, AlphasenseOPC-N2 dust sensor);

[0031] Industrial equipment operation monitoring module (current transformer, torque sensor);

[0032] Wearable physiological monitoring devices (Empatica E4 wristband, electromyography patches);

[0033] Edge computing gateway (-40℃~85℃ wide temperature range, IP67 protection rating);

[0034] Industrial vision equipment (Baslerace2 camera).

[0035] Preferably, the cross-domain collaboration module is compatible with Azure IoT and Alibaba Cloud ET Industrial Brain platform, supports Kubernetes cluster deployment to achieve elastic expansion, meets the concurrent access requirements of 2000+ sensors, and all data processing processes comply with the Personal Information Protection Law and ISO / IEC 27701 privacy standards.

[0036] The technical effects and advantages of this invention, a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, are as follows:

[0037] 1. This invention provides more comprehensive data dimensions. By linking multi-source data from the environment, equipment, and individuals, it breaks through the limitations of traditional single environmental parameters and improves the accuracy of risk assessment.

[0038] 2. This invention makes the evaluation model dynamic, achieves individual risk baseline adaptation based on federated learning, and accurately locates the risk source by combining causal inference, thus solving the problem of static threshold lag.

[0039] 3. This invention enables smarter management intervention, constructs a hierarchical linkage closed-loop mechanism, and links industrial equipment for rapid response, upgrading from passive alarm to proactive prevention and control, thereby reducing the incidence of risk events.

[0040] 4. This invention makes cross-domain collaboration safer by using privacy computing technology to achieve secure data sharing among multiple entities, connecting with medical and social security systems to form a collaborative occupational health management system that balances efficiency and compliance.

[0041] 5. This invention is adapted to the needs of industrial scenarios, deeply integrates industrial equipment operation and production process data, supports human-machine collaboration, digital twins and other scenarios, and balances occupational health and production efficiency. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet proposed in this invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0045] Example 1

[0046] refer to Figure 1 This embodiment provides a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, used for a multi-dimensional dynamic assessment system in chemical workshops. Specific implementation details include:

[0047] Technical solution:

[0048] Based on the occupational health management system for chemical workshops, it includes:

[0049] Multi-dimensional data acquisition module:

[0050] Environmental sensor deployment: Using the RockEMBISG-503 industrial gateway, a gridded environmental sensor network (SensirionSHT45 temperature and humidity sensor, Alphasense OPC-N2 dust sensor, CITYTECHNOLOGY electrochemical benzene series sensor) is deployed in the workshop, with 3 sets deployed per 100㎡ to cover the entire workshop area.

[0051] Personnel positioning and physiological monitoring: The deployment density of UWB positioning base stations (Decawave DW1000) is every 20m, with a positioning accuracy of ±0.1m; the wearable device uses the Empatica E4 wristband to collect 12 physiological indicators such as heart rate variability (HRV) and skin conductance (EDA) in real time, with a sampling frequency of 256Hz.

[0052] Path integral algorithm implementation:

[0053] formula: ),in The concentration of benzene series compounds in region i (unit: ppm) The dwell time is measured in minutes. Numerical integration was performed using the Gaussian quadrature method, with a time step of 1 second, to ensure that the cumulative exposure calculation error was less than 5%.

[0054] Multimodal data fusion: The industrial vision equipment (Baslerace2 series) uses the YOLOv8 algorithm to identify the wearing status of protective equipment and fuses it with environmental and physiological data to form a three-dimensional data matrix, which is updated once per second.

[0055] Dynamic evaluation model module:

[0056] Federated learning baseline optimization: Historical data from three plant areas (a total of 500,000 samples) were aggregated, and the Federated Averaging (FedAvg) algorithm was used for model training. Each aggregation round contained 10 local training cycles, for a total of 50 rounds. The dynamic baseline for young workers (<45 years old) was set at 150 ppm·h, and for older workers (≥45 years old) it was set at 100 ppm·h. The baseline difference rate between healthy workers and workers with a history of asthma reached 40%.

[0057] Causal analysis: A causal graph model was built using the Python CausalML library to identify that the risk of acute poisoning increased by 7 times when the benzene concentration was >80ppm and protective equipment was not worn (OR = 7.2, p < 0.001).

[0058] Intelligent management closed-loop module:

[0059] Three-tiered intervention mechanism:

[0060] Level 1 intervention: When the benzene concentration is >50ppm, a vibration and voice warning is sent to the worker's Empatica wristband via the LoRa module (SX1278), with a response time of <2 seconds.

[0061] Level 2 intervention: The Siemens PLC control system is linked to adjust the ventilation system frequency to 50Hz, and the reactor pressure is reduced to 0.8MPa via the OPCUA protocol. The execution time is less than 10 seconds.

[0062] Level 3 intervention: When an emergency evacuation order is triggered, the device is locked and emergency lighting is activated via the ModbusTCP protocol, and the evacuation route is displayed in real time via AR glasses (HoloLens2).

[0063] Effect verification: Siemens SimaticIPC industrial computers were deployed to monitor post-intervention data in real time and calculate the intervention effect index (EEI). Model iteration was triggered when EEI < 0.7, with an average iteration cycle of 2 hours.

[0064] Cross-domain collaboration module:

[0065] Privacy-preserving computing platform: Employs federated learning and homomorphic encryption (Paillier algorithm), ensuring that raw data does not leave the local node. For example, when calculating the average benzene exposure of three companies, the encrypted data from each company is processed using homomorphic addition, resulting in an error of <0.5ppm.

[0066] Social service linkage: It connects with the HIS system of the park's medical center. When the cumulative exposure of workers exceeds 200 ppm·h, it automatically triggers the appointment of occupational disease examination and ensures that the data is tamper-proof through blockchain (HyperledgerFabric).

[0067] Application scenarios

[0068] After deployment in the benzene workshop of a chemical company:

[0069] The number of benzene exposure incidents decreased from 12 per month to 3 (a 75% reduction), and the rate of acute poisoning incidents decreased from 0.5 per thousand man-hours to 0.08 per thousand man-hours (an 84% reduction).

[0070] Compared to the traditional fixed threshold (8-hour weighted average of 50ppm), the dynamic baseline improves the early warning accuracy from 58% to 91% and reduces the false alarm rate from 32% to 9%.

[0071] Example 2

[0072] This embodiment provides a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, used for human-machine collaboration risk prevention and control in automobile manufacturing. Specific implementation details include:

[0073] Technical solution:

[0074] A human-machine collaborative safety system includes:

[0075] Risk prediction model:

[0076] Data acquisition: The industrial robot (KUKAKR10R1420) transmits TCP position (accuracy ±0.02mm) and speed (resolution 0.1mm / s) in real time via the Profinet protocol; the UWB positioning base station (DecawaveDW1000) tracks the worker's position and integrates them to form a motion trajectory data stream.

[0077] LSTM Neural Network: Employs a 2-layer LSTM structure (100 neurons per layer) + fully connected layer, with an input dimension of 6 (robot TCP coordinates X / Y / Z, velocity Vx / Vy / Vz), and an output of collision probability (0-1). The training dataset contains 200,000 historical trajectories, using the Adam optimizer with a learning rate of 0.001, 100 training epochs, and achieving a test set accuracy of 95.3%.

[0078] Dynamic intervention mechanism:

[0079] Level 3 Protection Response:

[0080] When the probability is greater than 0.5: the robot switches to "compliant mode", the joint impedance is reduced by 50%, and it is adjusted in real time via EtherCAT protocol, with a response time of less than 50ms.

[0081] When the probability is greater than 0.7: an emergency stop command (STO safety function) is triggered. At the same time, HoloLens2 generates an avoidance path through the SLAM algorithm and projects it into the worker's field of vision. The response time is less than 200ms.

[0082] Dynamic delineation of safe zones: A risk heat map is generated based on a Gaussian mixture model (GMM). A radius of 1.5m around the welding station is set as a high-risk zone. When the robot enters this zone, it automatically slows down to 0.2m / s.

[0083] Data closed-loop optimization:

[0084] Digital twin mirroring: Using the Unity engine to build a virtual production line, the robot's movement status is mapped in real time, and a risk heat map is generated for each shift, marking high-risk periods (such as a 37% increase in risk 30 minutes before and after shift change).

[0085] Model iteration: Accumulated data training reduced the collision warning response time from 800ms to 200ms, and improved the model's generalization ability by 23%.

[0086] Application scenarios

[0087] After deployment in a certain automobile assembly workshop:

[0088] The human-machine collision accident rate decreased from 0.3 times / thousand man-hours to 0.05 times / thousand man-hours (a reduction of 83%), and the robot's operating efficiency increased by 18% due to the reduction in the safety protection radius.

[0089] The worker error rate dropped from 12% to 4%, and AR assistance reduced the completion time of complex processes by 27%.

[0090] Example 3

[0091] This embodiment provides a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, used for cross-domain collaborative management in steel enterprises. Specific implementation details include:

[0092] Technical solution:

[0093] Based on an industry-level collaborative system, including:

[0094] Privacy Computing Platform:

[0095] Federated Learning Architecture: Three steel companies collaborated on modeling using the Huawei Cloud Federated Learning Platform, employing a horizontal federated learning model. Each training round consisted of 5 local epochs, with a total of 100 iterations. Common industry risk factors were identified: when dust concentration >10mg / m³ and humidity >80%, the risk of pneumoconiosis increased threefold (HR=3.1, 95%CI: 2.5-3.9).

[0096] Homomorphic encryption application: The Paillier algorithm is used to encrypt the original data. For example, when calculating the average noise exposure value of 3 companies, the encrypted data of each company is processed by homomorphic addition, and the result error is <0.5dB.

[0097] Social service collaboration:

[0098] Healthcare system integration: Desensitized individual risk data is integrated with the hospital's PACS system via the FHIR protocol. When a worker's cumulative dust exposure exceeds 300 mg·h / m³, a low-dose CT screening appointment is automatically triggered.

[0099] Dynamic adjustment of social security: Integrating with the social security system API interface, adjusting the work injury insurance rate according to the worker's risk level, with a 20% increase for high-risk groups and a 15% decrease for low-risk groups.

[0100] Application scenarios

[0101] After application by a steel industry alliance:

[0102] The incidence of pneumoconiosis in the industry has decreased from 4.2 per 1,000 people per year to 3.0 per 1,000 people per year (a decrease of 29%), and the work injury insurance payout rate has decreased by 17%, saving more than 20 million yuan in compensation funds annually.

[0103] Cross-provincial data sharing complies with the requirements of the Personal Information Protection Law, reducing the risk of data leakage by 92%.

[0104] Example 4

[0105] This embodiment provides a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, used for digital twin simulation of electronic factories. Specific implementation details include:

[0106] Technical solution:

[0107] Based on a digital twin system, it includes:

[0108] Virtual factory image:

[0109] 3D modeling: Using Unreal Engine 5 to build a digital twin of the SMT workshop with an accuracy of 0.1mm, it can map 80% of the equipment status (pick-and-place machine, reflow oven, etc.) in real time.

[0110] Simulation analysis: Simulates the impact of different shift scheduling schemes (three-shift / two-shift) on occupational health, outputting indicators such as fatigue index (based on the NASA-TLX scale) and cumulative exposure. For example, a two-shift system reduces the worker's fatigue index by 22%, but increases dust exposure by 15%.

[0111] Augmented Reality Assistance:

[0112] Emergency drill: HoloLens2 overlays virtual evacuation routes and real-time risk heat maps, reducing worker evacuation time from 5 minutes to 90 seconds.

[0113] Equipment maintenance: Display a virtual protection area (e.g., 2m radius for a pick-and-place machine), and confirm operational compliance through gesture recognition (LeapMotion), reducing the maintenance error rate from 18% to 5%.

[0114] Application scenarios

[0115] After application by an electronics factory:

[0116] Optimizing the work schedule reduced worker fatigue index by 22% and occupational disease incidence by 19%.

[0117] Equipment maintenance efficiency increased by 35%, and emergency response time was shortened by 60%.

[0118] Example 5

[0119] This embodiment provides a multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, used for dynamic baseline optimization in new energy battery plants. Specific implementation details include:

[0120] Technical solution:

[0121] Based on a personalized assessment system, including:

[0122] Dynamic baseline generation:

[0123] Federated learning model: Data from 5 sites (totaling 1 million samples) were aggregated and optimized using the FedProx algorithm to distinguish the nickel exposure baseline between healthy workers and workers with a history of asthma. The baseline for healthy workers was 1.5 mg / m³ (8-hour weighted average), while the baseline for workers with a history of asthma was 0.8 mg / m³, with a difference rate of 46.7%.

[0124] Causal analysis: Using Do-Calculus to identify key factors: when nickel exposure >1 mg / m³ and ventilation system efficiency <80%, the risk of allergy increased 5-fold (OR=5.2, p<0.001).

[0125] Closed-loop intervention mechanism:

[0126] Intelligent adjustment: When a worker's HRV < 30ms and nickel exposure > 1.2 mg / m³, a secondary intervention is triggered.

[0127] Adjusting the welding robot parameters (reducing the current from 200A to 180A) reduced nickel volatilization by 23%.

[0128] By sending antihistamine medication reminders to workers' mobile phones via the MQTT protocol, medication adherence increased from 45% to 78%.

[0129] Model iteration: Accumulated data training improved the accuracy of allergy event prediction from 72% to 89%, and reduced the false alarm rate from 35% to 12%.

[0130] Application scenarios

[0131] After deployment at a lithium battery factory:

[0132] The incidence of allergic events decreased from 12% to 3.5% (a reduction of 71%), and the number of days off taken by workers decreased by 41%.

[0133] Compared to a fixed standard, the dynamic baseline reduces the false alarm rate by 58% and the equipment adjustment response time is less than 15 seconds.

[0134] Comparative Example 1

[0135] This comparison provides a traditional static threshold management system.

[0136] Technical solution

[0137] The occupational health system deployed by a chemical company in 2022 includes:

[0138] Data collection:

[0139] Only a single environmental sensor (Honeywell BWSolo gas detector) was deployed to monitor benzene concentration (fixed threshold 50ppm), without personnel location or physiological data.

[0140] Data is collected once per minute and stored on a local server; there is no data fusion function.

[0141] risk assessment:

[0142] A static threshold was used, without distinguishing between the age and health status of workers. For example, the same 50ppm threshold was used for older workers and younger workers.

[0143] The risk assessment period is 1 hour, which lags behind the actual exposure situation.

[0144] Intervention measures:

[0145] When the limit is exceeded, only an audible and visual alarm is triggered (85dB), without any equipment linkage or production adjustment. Workers need to manually shut down the equipment, with an average response time of >5 minutes.

[0146] Without effective validation and model iteration mechanisms, the accident rate has remained high for a long time.

[0147] Compared with Examples 1-5 and Comparative Example 1, the core difference of the present invention lies in the comprehensive upgrade of the technical architecture and actual effect:

[0148] In terms of data dimensions, the implementation example uses a three-dimensional matrix of environment, physiology, and behavior (such as 12 indicators in a chemical workshop and human-machine trajectory fusion in an automobile factory), while the comparative example only monitors a single environmental parameter (such as benzene concentration). The former has 11 times more data dimension coverage.

[0149] In terms of evaluation methods, the implementation example uses federated learning dynamic baselines (such as the differentiated threshold for new energy plants) and causal analysis (accuracy 91%), compared to a fixed threshold (accuracy 58%), dynamic evaluation reduces the false alarm rate from 32% to 9% and the false negative rate from 28% to 6%.

[0150] In terms of intervention mechanisms, the example constructs a three-level linkage closed loop (such as equipment control within 10 seconds in a chemical workshop and emergency stop within 0.2 seconds in an automobile factory), while the comparison only uses a single alarm + manual handling (response time exceeds 5 minutes), improving the response speed by more than 10 times.

[0151] In terms of data security, the implementation example uses federated learning and homomorphic encryption to reduce the risk of leakage to below 5%, while the comparative example has a 92% risk due to centralized storage of the original data, resulting in a significant improvement in compliance.

[0152] In terms of health benefits, the embodiment achieved a 29% reduction in the incidence of pneumoconiosis and a 71% reduction in the allergy rate, while the comparative embodiment showed no improvement; the device linkage supports 8 types of industrial protocols (such as OPCUA / Profinet), while the comparative embodiment only supports 1 type, which greatly improves compatibility.

[0153] In summary, the implementation method, through multi-dimensional integration and dynamic intelligent management, solves the core defects of the comparative method, which is static, singular, and lagging, and achieves a qualitative leap in occupational health management.

[0154] The above embodiments can be implemented in whole or in part by software, hardware, firmware or other arbitrary combinations. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0155] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0156] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0157] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0158] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional dynamic assessment and intelligent management system for occupational health risks based on the Industrial Internet, characterized in that: include: The multi-dimensional data acquisition module collects environmental parameters, industrial equipment operating parameters, and individual data of workers in real time through industrial Internet devices. It combines the data with the positioning system to calculate the real-time exposure of workers and build a multi-source data association mechanism for the environment, equipment, and people. The dynamic assessment model module constructs a dynamic assessment model based on multi-source data. It dynamically optimizes the individual risk baseline through a federated learning framework to address the lag in traditional static threshold assessment, uses causal inference algorithms to identify risk causal relationships, and achieves early prediction of risk trends through a time-series prediction model. The intelligent management closed-loop module links the industrial production control system and the resource scheduling system, executes hierarchical intervention measures based on dynamic evaluation results, and verifies and iterates the intervention effect in real time. The cross-domain collaboration module uses privacy computing technology to achieve secure sharing and collaborative analysis of data from multiple entities, and builds a collaborative occupational health management system covering enterprises, industries and social service systems.

2. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The multi-dimensional data acquisition module includes: The individual exposure quantification unit calculates the cumulative exposure of an individual in different areas based on the distribution data of the gridded environmental sensors in the workshop and the positioning trajectory of the workers through the path integral algorithm. It also dynamically adjusts the exposure threshold by associating individual physiological characteristic parameters, including age, medical history and real-time physiological status. The multimodal data fusion unit integrates physiological data collected by wearable devices, behavioral data collected by industrial vision devices, and environmental data collected by environmental sensors to form a three-dimensional data matrix of physiological state, behavioral norms, and environmental exposure intensity. The physiological data includes heart rate variability, skin conductance, and electromyography signals, while the behavioral data includes operating posture, protective equipment wearing status, and working duration.

3. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The dynamic evaluation model module includes: The Federated Learning Baseline Optimization Unit aggregates historical data from multiple plants through federated learning without sharing the original data, and dynamically updates the risk baselines of different individuals, thus differentiating the risk baselines of young workers from older workers, and healthy workers from workers with underlying medical conditions. The causal tracing unit analyzes the correlation between multi-source data through a causal graph model, distinguishes between the correlation and causality of risk factors, and accurately locates the core source of health risks. The core source includes environmental factors, equipment factors, or human factors.

4. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The intelligent management closed-loop module includes: The tiered intervention execution unit performs three levels of intervention based on the risk level. Level 1 intervention is to push early warning information to operators, Level 2 intervention is to coordinate with industrial equipment to adjust operating parameters, and Level 3 intervention is to trigger the suspension of production processes or the redistribution of work tasks. The effect verification iteration unit monitors changes in environmental parameters, equipment status, and physiological behavior data of workers in real time after intervention, calculates the intervention effect index, and feeds the index back to the dynamic evaluation model to optimize the model parameters.

5. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The cross-domain collaboration module includes: The privacy-protected data sharing unit employs federated learning and homomorphic encryption technologies to enable collaborative analysis of occupational health data across factories and enterprises, identifying common risk factors in the industry, without requiring the original data to leave the local machine. The social service linkage unit connects anonymized individual risk data with the medical and social security systems, triggering occupational health check-up reminders, medical diagnosis assistance, and dynamic adjustments to social security policies.

6. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The intelligent management closed-loop module also includes a digital twin sand table simulation unit, which is a virtual factory image based on the industrial internet. It simulates the distribution of occupational health risks under different production plans, shift schedules or equipment operating parameters, outputs optimized decision-making suggestions for balancing health risks and production efficiency, and uses augmented reality technology to overlay the virtual simulation results onto the actual production scenario to assist in emergency response.

7. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The intelligent management closed-loop module also includes a human-machine collaboration risk prevention and control unit, which integrates the motion trajectory data of the industrial robot with the real-time positioning data of the operator, predicts potential risks in the human-machine interaction process, dynamically adjusts the robot's operating parameters or triggers safety control commands, and pushes avoidance prompts to the operator.

8. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 2, characterized in that, The path integral algorithm of the individual exposure quantification unit is as follows: by associating the duration of the worker's stay in each area with the real-time concentration of environmental parameters in that area, the cumulative exposure of the individual is calculated. The environmental parameters include dust concentration, harmful gas concentration and noise decibel value.

9. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The industrial internet devices include environmental sensors, industrial equipment operation monitoring modules, wearable physiological monitoring devices, positioning base stations, industrial vision devices, industrial robot control systems, production execution systems, and logistics scheduling systems.

10. The occupational health risk multi-dimensional dynamic assessment and intelligent management system based on the Industrial Internet as described in claim 1, characterized in that, The privacy computing technology of the cross-domain collaboration module complies with industry data security standards. Through federated learning, it ensures that individual health data and enterprise production data do not leave the local node, but only share model parameters and analysis results.

Images