Wearable safety monitoring method, system and equipment for industrial operation and medium
By using a multimodal benchmark parameter library and dynamic network architecture to generate a domain adaptation model in an industrial operation environment, and combining LSTM, convolutional neural network and linear regression models for data analysis, the problems of inaccurate sensor data and insufficient adaptability of a single model are solved, and efficient and accurate safety risk assessment and classification are achieved.
Patent Information
- Application Number
- CN202510661383.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-22
- Publication Date
- 2025-09-12
AI Technical Summary
In industrial work environments, physiological and environmental parameters collected by sensors are susceptible to noise and outliers, resulting in inaccurate data and affecting the assessment of worker safety and working conditions. Furthermore, single models struggle to adapt to complex work environments and individual differences, making them incapable of accurately assessing and classifying safety risks.
A multimodal benchmark parameter library, dynamic network architecture, adaptive optimization strategy library and adversarial training mechanism are used to generate domain adaptation models, enhance the model's multi-task generalization capabilities, and analyze and process data through LSTM recurrent network models, convolutional neural network models and linear regression models to improve the accuracy of risk classification and assessment.
It improves the accuracy and efficiency of multimodal data risk classification, enhances the ability to judge workers' safety status and working environment conditions, ensures data quality and provides reliable safety decision-making support.
Smart Images
Figure CN120634221A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of industrial equipment monitoring, and in particular relates to a wearable safety monitoring method, system, equipment and medium for industrial operations. Background Art
[0002] With the continuous advancement of technology, the application of wearable devices, including smart bracelets and smart watches, is becoming increasingly widespread. In industrial work environments, the physiological and environmental parameters collected by sensors are susceptible to various interferences, such as noise and outliers. This leads to inaccurate data, affecting subsequent assessments of worker safety and working environment conditions. Monitoring with fixed sensors in work areas lacks effective data preprocessing, making it difficult to ensure data quality.
[0003] Industrial work scenarios are complex and diverse, with significant differences between different working environments and individual workers. A single model struggles to adapt to these complex situations and cannot accurately assess and classify worker safety risks. For example, in high-temperature and high-humidity environments, workers' physiological parameters change differently than in normal environments, and the model may not be able to accurately identify risks in such environments. Furthermore, when processing physiological and environmental parameters, the model struggles to fully exploit the correlations between the data, resulting in low accuracy in risk assessment and classification. Summary of the Invention
[0004] The present invention provides a wearable safety monitoring method for industrial operations. It utilizes a multimodal benchmark parameter library, a dynamic network architecture, an adaptive optimization strategy library, and an adversarial training mechanism to generate a domain adaptation model, thereby enhancing the model's multi-task generalization capability and improving the accuracy and efficiency of multimodal data risk classification.
[0005] Methods include: Collect physiological and environmental parameters through parameter monitoring equipment; remove abnormal values and perform smoothing preprocessing on the collected physiological and environmental parameters; Configure the processed physiological and environmental parameters with a timestamp and upload them to the cloud; Obtaining original physiological and environmental parameters including timestamps from the cloud, preprocessing the original physiological and environmental parameters to obtain processed physiological and environmental parameters; Based on the preset LSTM recurrent network model, physiological and environmental parameters are analyzed and processed to obtain corresponding spatiotemporal trend analysis results; Based on the preset convolutional neural network model, physiological and environmental parameters are analyzed and processed to obtain corresponding multimodal risk classification output information; Analyze and process physiological and environmental parameters based on the preset linear regression model to obtain corresponding confidence risk assessment information; The spatiotemporal trend analysis results, multimodal risk classification output information and confidence risk assessment information are integrated and processed to obtain safety decision output information.
[0006] Preferably, the step of preprocessing the original physiological and environmental parameters to obtain the processed physiological and environmental parameters specifically includes: Obtain preset data cleansing solutions and standardized mapping rules; Perform data purification on the original physiological and environmental parameters based on the data purification scheme to obtain an intermediate data set; Performing structured mapping processing on the intermediate data set based on standardized mapping rules to obtain a structured data set; The structured data sets are used as the final processed physiological and environmental parameters.
[0007] Preferably, before the step of analyzing and processing the physiological and environmental parameters based on a preset linear regression model to obtain corresponding confidence risk assessment information, the method further includes: Acquiring a pre-built multi-source heterogeneous parameter set, wherein the set includes a cross-modal dataset of physiological parameters and environmental parameters; Select an appropriate probabilistic graph model based on the research purpose; Loading dynamic parameter optimization protocol, which integrates variational inference rules and Markov chain Monte Carlo sampling strategy; Based on the dynamic parameter optimization protocol, the Bayesian hierarchical optimization framework is used to perform hyperparameter tuning and model verification on the probabilistic graph model, generating a probabilistic linear regression engine that meets the uncertainty quantification requirements. Embeds a probabilistic linear regression engine into a linear regression model.
[0008] Preferably, before the step of analyzing and processing the physiological and environmental parameters based on the preset convolutional neural network model to obtain corresponding multimodal risk classification output information, the method further includes: Obtaining a pre-built multimodal baseline parameter library, the parameter library containing a historical sample set of worker physiological parameters and environmental parameters; Calling a basic model framework and determining a dynamic network architecture that matches the framework; Loading an adaptive optimization strategy library, wherein the strategy library includes transfer learning rules and dynamic pruning rules; Based on the adaptive optimization strategy library, an adversarial training mechanism is used to optimize the parameters and verify the performance of the basic model framework, generating a domain adaptation model that meets the generalization requirements of multiple tasks; The domain adaptation model is deployed as a convolutional neural network model.
[0009] Preferably, after the step of outputting the security decision output information, the method further includes: obtaining a preset dynamic adjustment strategy; Call preset physiological and environmental analysis models; Construct a safety update strategy corresponding to the physiological and environmental analysis model based on the dynamic adjustment strategy and safety decision output information; Physiological and environmental analysis models are updated based on the security update strategy.
[0010] Preferably, the method further comprises a calculation method for obtaining a preset physiological parameter; Get the preset calculation adjustment strategy; Based on the safety decision output information, the calculation method of the physiological parameters is adjusted accordingly using the calculation adjustment strategy.
[0011] Preferably, the method also monitors the safety decision output information in real time, issues warning information when abnormal situations are found, and notifies on-site personnel; The received physiological and environmental parameters are also summarized to form a database with partitioned storage of physiological parameters and work environment parameters. The Spark streaming computing engine is used to realize multi-device data fusion analysis, and the 3D-CNN deep learning algorithm is used to predict the risks of falls and suffocation, and generate real-time reports including workers' physiological status and environmental safety status. The present application also provides a wearable safety monitoring system for industrial operations, the system comprising: The parameter acquisition module is used to collect physiological and environmental parameters through parameter monitoring equipment; remove abnormal values and smooth preprocessing of the collected physiological and environmental parameters; The upload module is used to configure the timestamp of the processed physiological and environmental parameters and upload them to the cloud; A preprocessing module is used to obtain the original physiological and environmental parameters including timestamps from the cloud, preprocess the original physiological and environmental parameters, and obtain processed physiological and environmental parameters; The time-controlled analysis module is used to analyze and process physiological and environmental parameters in combination with the preset LSTM recurrent network model to obtain corresponding spatiotemporal trend analysis results; The risk classification module is used to analyze and process physiological and environmental parameters using a preset convolutional neural network model to obtain corresponding multimodal risk classification output information; The risk assessment module is used to analyze and process physiological and environmental parameters using a preset linear regression model to obtain corresponding confidence risk assessment information; The integration processing module is used to integrate the spatiotemporal trend analysis results, multimodal risk classification output information and confidence risk assessment information to obtain safety decision output information.
[0012] According to another embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the wearable safety monitoring method for industrial operations when executing the program.
[0013] According to another embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the wearable safety monitoring method for industrial operations are implemented.
[0014] It can be seen from the above technical solutions that the present invention has the following advantages: This application proposes a wearable safety monitoring method for industrial operations. This method uses parameter monitoring equipment to collect physiological and environmental parameters and performs multiple preprocessing steps to ensure data accuracy and reliability. By combining LSTM recurrent network models, convolutional neural network models, and linear regression models, the method analyzes data from multiple perspectives, including spatiotemporal trends, multimodal risk classification, and confidence risk assessment, enabling the understanding of workers' safety status and working environment conditions.
[0015] Before using the linear regression model, a multi-source heterogeneous parameter set is constructed, a probabilistic graph model is selected, and hyperparameters are tuned through a dynamic parameter optimization protocol to generate a probabilistic linear regression engine. This improves the linear regression model's ability to quantify uncertainty and enables more accurate risk assessment. Before using the convolutional neural network model, a domain-adapted model is generated using a multimodal benchmark parameter library, a dynamic network architecture, an adaptive optimization strategy library, and an adversarial training mechanism. This enhances the model's multi-task generalization capabilities and improves the accuracy and efficiency of multimodal data risk classification. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0017] Figure 1 A flow chart of a wearable safety monitoring method for industrial operations; Figure 2 Schematic diagram of a wearable safety monitoring system for industrial operations; Figure 3 Schematic diagram of an electronic device. DETAILED DESCRIPTION
[0018] The wearable safety monitoring method for industrial operations provided in this application integrates multimodal sensors and intelligent algorithms to monitor workers' heart rate, body temperature, and environmental harmful gases, temperature, and humidity in real time. Combined with a dynamic early warning system and cloud-based data synchronization, it enhances proactive safety protection capabilities in industrial scenarios. Risks are reduced through an abnormal state response mechanism, while historical data tracing provides a reliable basis for optimizing safety management and determining accident responsibility, ensuring personnel safety.
[0019] The following describes in detail the wearable safety monitoring method for industrial operations involved in this application. For the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are provided to facilitate a thorough understanding of the embodiments of this application. However, it should be clear to those skilled in the art that this application can also be implemented in other embodiments without these specific details.
[0020] It should be understood that when used in this specification, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their collections. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.
[0021] The phrases "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of the application. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.
[0022] The wearable safety monitoring method for industrial operations provided in this application can acquire and process relevant data based on artificial intelligence technology. Specifically, the wearable safety monitoring method for industrial operations utilizes digital computer-controlled machines to simulate, extend, and expand human intelligence, and provides theories, methods, techniques, and application devices for perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.
[0023] The wearable safety monitoring method for industrial operations also has a machine learning function, wherein the machine learning and deep learning in the method of the present invention generally include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.
[0024] In embodiments of the present invention, computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including, but not limited to, object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (exemplarily, via the Internet using an Internet service provider).
[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0026] See also Figure 1 FIG. 1 is a flow chart of a wearable safety monitoring method for industrial operations according to a specific embodiment. The method includes: S101: Collect physiological and environmental parameters through parameter monitoring equipment; remove outliers and perform smoothing preprocessing on the collected physiological and environmental parameters; physiological and environmental parameters include: physiological parameters such as heart rate, blood pressure, respiratory rate, body temperature, as well as parameters such as temperature, humidity, harmful gas concentration, and noise in the environment.
[0027] The wearable device of this embodiment has a built-in multimodal sensor array, such as a PPG optical heart rate sensor, a MEMS barometer, and a thermopile body temperature sensor, which synchronously collects data streams via the SPI bus. The Grubbs test is used to dynamically scan outliers, combined with the isolation forest algorithm to identify systematic deviations caused by sensor drift. For example, when the heart rate is >200bpm or <30bpm, the three-level verification mechanism is triggered. Wavelet transform denoising is performed on the time series data to filter out high-frequency noise in the frequency domain and retain the characteristics of physiological signals. The timestamp alignment algorithm is used to unify the sampling frequencies of multiple sensors, such as 1Hz for physiological parameters and 10Hz for environmental parameters.
[0028] S102: The processed physiological and environmental parameters are timestamped and uploaded to the cloud.
[0029] In this embodiment, each physiological and environmental parameter is attached with a timestamp and a device ID hash value and uploaded to the cloud storage via the MQTT protocol.
[0030] The timestamp usually adopts a standard time format. Before uploading the physiological and environmental parameters, the physiological and environmental parameters are encrypted to ensure the security of the data during transmission. Symmetric encryption algorithms such as AES algorithm can be used.
[0031] S103: Obtaining original physiological and environmental parameters including timestamps from the cloud, preprocessing the original physiological and environmental parameters, and obtaining processed physiological and environmental parameters.
[0032] This embodiment obtains the original physiological and environmental parameters including timestamps from the cloud server. Since the data stored in the cloud may have new noise or anomalies during transmission or storage, it needs to be pre-processed again.
[0033] As an implementation of this embodiment, a preset data cleansing scheme and standardized mapping rules are obtained. The data cleansing scheme defines how to handle missing values, outliers, and noisy data, such as sensor drift of physiological parameters. The standardized mapping rules define how to unify data into a standardized format, such as converting sensor units from "mmHg" to "kPa" or unifying timestamp formats.
[0034] Data cleansing is performed on the original physiological and environmental parameters based on the data cleansing solution to generate an intermediate dataset. This data cleansing process includes filling missing values, removing outliers, and filtering noise. The intermediate dataset is the cleaned data, preserving the original data structure but removing invalid or anomalous components.
[0035] Structured mapping is performed on the intermediate dataset based on standardized mapping rules to produce a structured dataset. Structured mapping converts data into a unified format. This standardized dataset meets the requirements of subsequent model inputs, such as aligning multi-source sensor data to a common time base.
[0036] The structured dataset is used as the final processed physiological and environmental parameters. The final output is clearly a structured dataset that meets the model input requirements.
[0037] S104: Analyze and process physiological and environmental parameters based on a preset LSTM recurrent network model to obtain corresponding spatiotemporal trend analysis results.
[0038] This example uses an LSTM recurrent network model to process sequential data and capture the temporal dependencies between physiological and environmental parameters. Using physiological and environmental parameters as input sequences, the LSTM model learns patterns and trends within the sequences to predict future data changes, resulting in spatiotemporal trend analysis results.
[0039] It's important to note that the LSTM model consists of an input layer, a hidden layer, and an output layer. The memory cells in the hidden layer can store long-term information, avoiding the vanishing gradient problem found in traditional recurrent neural networks. Training an LSTM model requires a large amount of historical data, and the model's weight parameters are continuously adjusted using a backpropagation algorithm.
[0040] S105: Analyze and process physiological and environmental parameters based on a preset convolutional neural network model to obtain corresponding multimodal risk classification output information.
[0041] The convolutional neural network model of this embodiment can automatically learn the characteristic patterns in physiological and environmental parameters, fuse and analyze physiological parameters and environmental parameters of different modalities, and classify risks.
[0042] It should be noted that the convolutional neural network model consists of convolutional layers, pooling layers, and fully connected layers. The convolutional layers extract local features of the data using convolution kernels, the pooling layers perform dimensionality reduction on these features, and the fully connected layers map the extracted features to different risk categories.
[0043] For example, multimodal fusion combines heart rate, respiratory rate, hazardous gas concentrations, and noise into multi-channel inputs. Key features are extracted through convolutional layers, for example, detecting areas where CO concentration exceeds a threshold. This correlates and identifies high-temperature-induced shortness of breath and elevated heart rate. The output risk categories include high suffocation risk, heatstroke warning, and a confidence level, such as a 90% confidence level for high-temperature risk.
[0044] S106: Analyze and process physiological and environmental parameters based on a preset linear regression model to obtain corresponding confidence risk assessment information. The linear regression model of this embodiment establishes a linear relationship between physiological and environmental parameters and risk, and predicts the confidence level of risk based on the input parameters. Assuming a linear relationship between risk and parameters, the model coefficients are determined by minimizing the error between the predicted value and the actual value.
[0045] S107: Integrate and process the spatiotemporal trend analysis results, multimodal risk classification output information, and confidence risk assessment information to obtain safety decision output information.
[0046] This embodiment integrates spatiotemporal trend analysis results, multimodal risk classification output, and confidence risk assessment information to comprehensively consider various risk factors. This information integration can be achieved by using a weighted average approach, assigning weights based on the importance of different pieces of information. Alternatively, machine learning algorithms such as decision trees can be used to make decisions based on the combination of different pieces of information.
[0047] Optionally, this embodiment converts the LSTM trend prediction confidence level of 0.7, the CNN classification result confidence level of 0.85, and the BLR risk assessment confidence level of 0.9 into a basic probability distribution function, and calculates the final risk level using the Dempster combination rule.
[0048] If the conflict factor K is less than 0.2, the overall risk level is output. If K is greater than or equal to 0.2, a manual review process is triggered. The decision output of this embodiment includes a three-dimensional risk heat map, i.e., a three-dimensional risk heat map based on time, space, and risk type. A vibration alert can be sent to a wearable device.
[0049] Based on the above embodiment, before the step of analyzing and processing the physiological and environmental parameters based on the preset linear regression model to obtain corresponding confidence risk assessment information, the following steps are further included: A pre-built multi-source heterogeneous parameter set is obtained, wherein the set includes a cross-modal dataset of physiological parameters and environmental parameters.
[0050] Here, workers' physiological parameters and work environment parameters are collected through different types of sensors. These data vary in source, format, and characteristics, forming multi-source heterogeneous data. This data is integrated to construct a cross-modal dataset that includes both physiological and environmental parameters.
[0051] Select an appropriate probabilistic graph model based on the research purpose; A probabilistic graphical model is a mathematical model that uses a graph to represent the probabilistic dependencies between random variables. Risk assessment studies based on physiological and environmental parameters require analysis of causal or interdependent relationships between these parameters. For example, if the goal is to analyze the impact of different environmental factors on a specific physiological indicator of a worker, a Bayesian network can clearly represent the causal relationships between variables using a directed acyclic graph. Markov random fields, on the other hand, are more suitable for handling undirected dependencies between variables. When selecting a probabilistic graphical model, the appropriate structure should be determined based on the actual relationship between the parameters and the research objectives, along with the definition of the corresponding random variables and probability distributions.
[0052] Loading dynamic parameter optimization protocol, which integrates variational inference rules and Markov chain Monte Carlo sampling strategy; The dynamic parameter optimization protocol of this embodiment is designed to optimize the parameters in a probabilistic graph model. The Markov Chain Monte Carlo sampling strategy constructs a Markov chain and converges it to a target distribution, thereby obtaining samples for parameter estimation. The protocol integrates these two methods. First, based on the characteristics of the probabilistic graph model and the data, the approximate distribution form of variational inference and the initial state of the Markov chain are determined. During the optimization process, variational inference is used to quickly obtain approximate parameter estimates. These estimates are then refined and adjusted through Markov Chain Monte Carlo sampling, continuously updating the parameter values to bring them closer to the true optimal parameters.
[0053] Based on the dynamic parameter optimization protocol, a Bayesian hierarchical optimization framework is used to perform hyperparameter tuning and model validation on the probabilistic graph model, generating a probabilistic linear regression engine that meets uncertainty quantification requirements. The Bayesian hierarchical optimization framework divides the parameters of the probabilistic graph model into multiple layers, taking into account the hierarchical dependencies between parameters. Within this framework, the dynamic parameter optimization protocol is used to adjust the hyperparameters of the probabilistic graph model. Hyperparameters control the model structure and parameter distribution, such as the variance of the probability distribution. By continuously adjusting hyperparameters, the model achieves optimal performance on the training data.
[0054] Embeds a probabilistic linear regression engine into a linear regression model.
[0055] The linear regression model of this embodiment is a model that makes predictions based on the linear relationship between independent and dependent variables. Embedding a probabilistic linear regression engine into it replaces the parameter estimation and prediction components of the original linear regression model with the probabilistic linear regression engine. When physiological and environmental parameters are input, the probabilistic linear regression engine uses its optimized and validated parameters and algorithms to perform probability-based linear regression calculations, outputting confidence risk assessment information that includes quantified uncertainty. This can better meet the needs for accurate risk assessment and comprehensive understanding in practical applications.
[0056] Before the step of analyzing and processing physiological and environmental parameters based on a preset convolutional neural network model to obtain corresponding multimodal risk classification output information, this embodiment also includes: obtaining a pre-built multimodal benchmark parameter library, which contains a historical sample set of workers' physiological parameters and environmental parameters.
[0057] The parameter library of this embodiment forms a rich historical sample set by collecting workers' heart rate, blood pressure, respiratory rate, body temperature, temperature, humidity, harmful gas concentration, noise, etc. in different work scenarios over a long period of time.
[0058] The base model framework is called and the dynamic network architecture that matches the framework is determined.
[0059] Dynamic network architectures flexibly adjust and optimize the basic model framework based on specific task requirements and data characteristics. When determining a dynamic network architecture, factors such as the input data dimension, number of channels, and inter-data correlation need to be considered. Furthermore, by adjusting parameters such as the number of network layers, convolution kernel size, and pooling methods, a network structure optimally suited for processing multimodal physiological and environmental parameter data can be constructed.
[0060] An adaptive optimization strategy library is loaded, wherein the strategy library includes transfer learning rules and dynamic pruning rules.
[0061] Based on the adaptive optimization strategy library, an adversarial training mechanism is used to optimize the parameters and verify the performance of the basic model framework to generate a domain adaptation model that meets the generalization requirements of multiple tasks.
[0062] The adversarial training mechanism introduces an adversarial network, consisting of a generator and a discriminator. The generator attempts to generate samples that are similar to the real data distribution in order to deceive the discriminator; the discriminator strives to distinguish between the generated samples and the real samples. This embodiment of the mechanism can enhance the model's understanding and adaptability to data distribution, improve the model's robustness, and enable it to maintain good performance in the face of noise, data changes, and other situations. By combining an adaptive optimization strategy, while optimizing model parameters, the complexity of the model is reduced to meet the generalization requirements in multi-task scenarios of industrial operations.
[0063] The domain adaptation model, optimized and verified through the above steps, is packaged and configured according to the standard structure and operating mode of a convolutional neural network. This includes setting up input and output interfaces, data processing procedures, and model parameter storage and loading methods to enable integration with the actual monitoring system. During deployment, the model is adaptively adjusted based on the actual hardware environment and software platform to ensure stable operation and accurate output of multimodal risk classification results.
[0064] In some specific embodiments, after the step of outputting the safety decision output information, the method further includes: obtaining a preset dynamic adjustment strategy; and calling a preset physiological and environmental analysis model.
[0065] Dynamic adjustment strategies are pre-defined rules based on the characteristics of industrial operations, safety management requirements, and past experience. These rules account for a variety of influencing factors, such as operation type, individual worker differences, and environmental dynamics. They operate by defining a series of conditions and corresponding adjustments to address changing safety monitoring requirements in different situations.
[0066] The physiological and environmental analysis models in this embodiment have undergone preliminary training and validation, enabling effective analysis of physiological and environmental parameters. These models are built based on machine learning or deep learning algorithms, such as the LSTM recurrent network model and convolutional neural network model mentioned above. When a model is called, the system loads the model structure and pre-trained parameters. They operate by using the model structure and parameters to extract features, identify patterns, and perform analysis and prediction on input data.
[0067] Based on the dynamic adjustment strategy and safety decision output information, a safety update strategy corresponding to the physiological and environmental analysis model is constructed. The dynamic adjustment strategy can be combined with the safety decision output information to analyze the safety status reflected by the current safety decision output information and the adjustment requirements of the dynamic adjustment strategy to address this situation. A specific update strategy can be developed based on the characteristics and operating principles of the physiological and environmental analysis model. For example, if the safety decision output information indicates low prediction accuracy for a certain risk, and the dynamic adjustment strategy requires improving prediction capabilities, then an update strategy for the convolutional neural network model may include increasing the amount of training data or adjusting network structure parameters.
[0068] Physiological and environmental analysis models are updated based on a safety update strategy. Specific updates can be performed on the physiological and environmental analysis models according to the established safety update strategy. To add training data, the new data is added to the training set and the model is retrained. To adjust network structure parameters, the model's configuration is modified and retrained. The working principle is to use the new strategy and data to optimize the model's parameters or structure, enabling it to better fit the data and predict safety conditions.
[0069] This embodiment can also obtain a preset calculation adjustment strategy, which is a set of rules and methods for adjusting the physiological parameter calculation method in order to cope with different scenarios and needs. Various factors are taken into account, such as the accuracy changes of the measuring equipment, the impact of individual differences of workers on the measurement results, and the change patterns of physiological parameters under different working environments. Based on the safety decision output information, the calculation adjustment strategy is used to make corresponding adjustments to the physiological parameter calculation method. This is to judge whether the current physiological parameter calculation method meets the needs of safety monitoring based on the safety decision output information. If the safety decision output information shows that the analysis results of certain physiological parameters have deviations or large uncertainties, the corresponding physiological parameter calculation method is adjusted in combination with the calculation adjustment strategy.
[0070] If blood pressure measurement results are found to fluctuate significantly under certain operating conditions, impacting safety decisions, the calculation adjustment strategy may include adjusting the filter parameters in the blood pressure calculation method or adopting a new calculation model. This can better adapt to different situations, ensure the accuracy of physiological parameters, and provide a more reliable basis for safety decisions.
[0071] As an embodiment of the present application, a monitoring center can also be configured. After receiving the data, the monitoring center monitors the data in real time. When an abnormal situation is found, early warning information is immediately issued to notify on-site managers and workers to take corresponding countermeasures. The received data is aggregated to form a worker's physiological parameter database and a work environment database, and the data in the database is deeply analyzed. The Spark streaming computing engine is used to realize multi-device data fusion analysis, and the 3D-CNN deep learning algorithm is used to predict risks such as falls and suffocation. Real-time reports containing workers' physiological status, environmental safety status, etc. are generated to provide decision support for on-site managers. The monitoring center supports querying historical data, analyzing workers' long-term work status and environmental change trends, building three-dimensional work scenarios based on the industrial digital twin platform, and linking the emergency broadcast system to initiate evacuation plans.
[0072] The following is a specific example to illustrate the implementation process of this method: a factory uses smart bracelets to monitor worker safety and prevent the risk of poisoning from high temperature and harmful gases.
[0073] Workers wear smart wristbands that collect real-time physiological parameters including heart rate (70-100 bpm), body temperature (36-37°C), and respiratory rate (12-20 breaths / minute). Environmental parameters include ambient temperature (initial 30°C), humidity (60%), and CO concentration (0-50 ppm).
[0074] If the CO concentration suddenly rises to 200 ppm at a certain moment, it is marked as an outlier and removed. Smoothing is then performed, applying a 5-point moving average to the temperature data to eliminate ±1°C fluctuations caused by sensor jitter. The data is timestamped, such as "2025-04-29 08:30:00," and encrypted and transmitted to the cloud via the MQTT protocol. The CO concentration data is converted from ppm to mg / m³, and missing humidity data is linearly interpolated.
[0075] Based on the preset LSTM recurrent network model, the temperature data for the past two hours (30°C → 35°C → 38°C) is input. The LSTM recurrent network model predicts that the temperature will rise to 40°C in the next hour. A sudden temperature change is detected, such as a 5°C increase in 30 minutes.
[0076] The convolutional neural network model captured physiological parameters: heart rate 100 bpm (upper limit of normal), respiratory rate 25 breaths / minute (abnormal). Environmental parameters: CO concentration 35 ppm (above the safety threshold), temperature 38°C. Output: Risk classification: Combined risk of high temperature and CO poisoning. Confidence level: 95%.
[0077] The linear regression model calculated the risk index: Risk Index = 0.2 + 0.3 × (38-30) + 0.4 × (35 / 50) + 0.1 × (25-20) = 8.7 / 10. Confidence Interval: Risk Index 8.7 ± 0.2. The spatiotemporal trend analysis results, multimodal risk classification output, and confidence risk assessment information were integrated and processed. The LSTM predicted persistent high temperatures, the CNN classified high risk, and the linear regression index was > 8.5. The wristband vibrated and a cloud-based notification prompted immediate evacuation of the high-temperature area. The decision was recorded based on a predicted temperature of 40°C and a risk of 95%. Ultimately, the workers received an early warning 15 minutes before the CO concentration exceeded the standard, allowing them to evacuate promptly and avoid poisoning. The system also recorded the complete data chain for subsequent safety analysis.
[0078] The following is an embodiment of a wearable safety monitoring system for industrial operations provided by an embodiment of the present disclosure. This system and the wearable safety monitoring methods for industrial operations of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the wearable safety monitoring system for industrial operations, please refer to the embodiment of the above-mentioned wearable safety monitoring method for industrial operations.
[0079] The system includes: a parameter acquisition module, which is used to collect physiological and environmental parameters through parameter monitoring equipment; and remove abnormal values and smooth preprocessing on the collected physiological and environmental parameters.
[0080] The upload module is used to configure the timestamp of the processed physiological and environmental parameters and upload them to the cloud.
[0081] The preprocessing module is used to obtain the original physiological and environmental parameters including timestamps from the cloud, preprocess the original physiological and environmental parameters, and obtain processed physiological and environmental parameters.
[0082] The time control analysis module is used to analyze and process physiological and environmental parameters in combination with the preset LSTM recurrent network model to obtain corresponding spatiotemporal trend analysis results.
[0083] The risk classification module is used to analyze and process physiological and environmental parameters using a preset convolutional neural network model to obtain corresponding multimodal risk classification output information.
[0084] The risk assessment module is used to analyze and process physiological and environmental parameters using a preset linear regression model to obtain corresponding confidence risk assessment information.
[0085] The integration processing module is used to integrate the spatiotemporal trend analysis results, multimodal risk classification output information and confidence risk assessment information to obtain safety decision output information.
[0086] like Figure 3As shown, the present application also provides an electronic device, including a display module 103, a memory 102, a processor 101, and a computer program stored in the memory and executable on the processor 101, wherein the processor 101 implements the steps of a wearable safety monitoring method for industrial operations when executing the program.
[0087] In the embodiments of the present invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.
[0088] In the embodiment of the present application, the processor 101 can be implemented by using at least one of a special purpose integrated circuit, a programmable logic device, a field programmable gate array, a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an embodiment can be implemented in a controller. For software implementation, an embodiment such as a process or function can be implemented with a separate software module that allows the execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.
[0089] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, which may be configured in the form of a liquid crystal display, an organic light emitting diode, etc.
[0090] The memory 102 can be used to store software programs and various data. The memory 102 can include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0091] The present application also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the wearable safety monitoring method for industrial operations.
[0092] The storage medium can be any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
[0093] In the context of storage media, a readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.
[0094] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A wearable safety monitoring method for industrial operations, characterized in that: Methods include: Collect physiological and environmental parameters through parameter monitoring equipment; Perform outlier removal and smoothing preprocessing on the collected physiological and environmental parameters; Configure the processed physiological and environmental parameters with a timestamp and upload them to the cloud; Obtaining original physiological and environmental parameters including timestamps from the cloud, preprocessing the original physiological and environmental parameters to obtain processed physiological and environmental parameters; Based on the preset LSTM recurrent network model, physiological and environmental parameters are analyzed and processed to obtain corresponding spatiotemporal trend analysis results; Based on the preset convolutional neural network model, physiological and environmental parameters are analyzed and processed to obtain corresponding multimodal risk classification output information; Analyze and process physiological and environmental parameters based on the preset linear regression model to obtain corresponding confidence risk assessment information; The spatiotemporal trend analysis results, multimodal risk classification output information and confidence risk assessment information are integrated and processed to obtain safety decision output information.
2. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: The step of preprocessing the original physiological and environmental parameters to obtain the processed physiological and environmental parameters specifically includes: Obtain preset data cleansing solutions and standardized mapping rules; Perform data purification on the original physiological and environmental parameters based on the data purification scheme to obtain an intermediate data set; Performing structured mapping processing on the intermediate data set based on standardized mapping rules to obtain a structured data set; The structured data sets are used as the final processed physiological and environmental parameters.
3. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: Before the step of analyzing and processing the physiological and environmental parameters based on the preset linear regression model to obtain corresponding confidence risk assessment information, the method further includes: Acquiring a pre-built multi-source heterogeneous parameter set, wherein the set includes a cross-modal dataset of physiological parameters and environmental parameters; Select an appropriate probabilistic graph model based on the research purpose; Loading a dynamic parameter optimization protocol that integrates variational inference rules and Markov chain Monte Carlo sampling strategies; Based on the dynamic parameter optimization protocol, the Bayesian hierarchical optimization framework is used to perform hyperparameter tuning and model verification on the probabilistic graph model, generating a probabilistic linear regression engine that meets the uncertainty quantification requirements. Embeds a probabilistic linear regression engine into a linear regression model.
4. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: Before the step of analyzing and processing the physiological and environmental parameters based on the preset convolutional neural network model to obtain corresponding multimodal risk classification output information, the method further includes: Obtaining a pre-built multimodal baseline parameter library, the parameter library containing a historical sample set of worker physiological parameters and environmental parameters; Calling a basic model framework and determining a dynamic network architecture that matches the framework; Loading an adaptive optimization strategy library, wherein the strategy library includes transfer learning rules and dynamic pruning rules; Based on the adaptive optimization strategy library, an adversarial training mechanism is used to optimize the parameters and verify the performance of the basic model framework, generating a domain adaptation model that meets the generalization requirements of multiple tasks; The domain adaptation model is deployed as a convolutional neural network model.
5. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: After the step of outputting the security decision output information, the method further includes: obtaining a preset dynamic adjustment strategy; Call preset physiological and environmental analysis models; Construct a safety update strategy corresponding to the physiological and environmental analysis model based on the dynamic adjustment strategy and safety decision output information; Physiological and environmental analysis models are updated based on the security update strategy.
6. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: The method also includes a calculation method for obtaining a preset physiological parameter; Get the preset calculation adjustment strategy; Based on the safety decision output information, the calculation method of the physiological parameters is adjusted accordingly using the calculation adjustment strategy.
7. The wearable safety monitoring method for industrial operations according to claim 1, characterized in that: The method also monitors the safety decision output information in real time and issues warning information to notify on-site personnel when abnormal situations are found; The received physiological and environmental parameters are also summarized to form a database with partitioned storage of physiological parameters and work environment parameters. The Spark streaming computing engine is used to realize multi-device data fusion analysis, and the 3D-CNN deep learning algorithm is used to predict the risks of falls and suffocation, and generate real-time reports including workers' physiological status and environmental safety status.
8. A wearable safety monitoring system for industrial operations, characterized in that: The system is used to implement the wearable safety monitoring method for industrial operations as described in any one of claims 1 to 7; The system includes: The parameter acquisition module is used to collect physiological and environmental parameters through parameter monitoring equipment; remove abnormal values and smooth preprocessing of the collected physiological and environmental parameters; The upload module is used to configure the timestamp of the processed physiological and environmental parameters and upload them to the cloud; A preprocessing module is used to obtain the original physiological and environmental parameters including timestamps from the cloud, preprocess the original physiological and environmental parameters, and obtain processed physiological and environmental parameters; The time-controlled analysis module is used to analyze and process physiological and environmental parameters in combination with the preset LSTM recurrent network model to obtain corresponding spatiotemporal trend analysis results; The risk classification module is used to analyze and process physiological and environmental parameters using a preset convolutional neural network model to obtain corresponding multimodal risk classification output information; The risk assessment module is used to analyze and process physiological and environmental parameters using a preset linear regression model to obtain corresponding confidence risk assessment information; The integration processing module is used to integrate the spatiotemporal trend analysis results, multimodal risk classification output information and confidence risk assessment information to obtain safety decision output information.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the wearable safety monitoring method for industrial operations as described in any one of claims 1 to 7 are implemented.
10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wearable safety monitoring method for industrial operations as claimed in any one of claims 1 to 7 are implemented.