Intelligent production safety inspection and early warning method and system fusing internet of things data

Abstract

The application relates to a safety production intelligent inspection and early warning method and system fusing Internet of Things data, the method comprising the following steps: acquiring multi-source Internet of Things monitoring data; performing edge side preprocessing on the multi-source Internet of Things monitoring data to obtain preprocessed monitoring data; extracting multi-modal state features based on the preprocessed monitoring data, constructing a device operation state feature set, and performing space-time alignment processing on the device operation state feature set to obtain fusion feature data in a unified reference coordinate; inputting the fusion feature data into a pre-constructed abnormality recognition model to perform health state evaluation and obtain a device operation deviation degree result; performing hierarchical early warning judgment according to the device operation deviation degree result to obtain an early warning level, and triggering a corresponding early warning disposal strategy according to the early warning level. The application has the effect of improving the accuracy of safety production inspection.

Description

Technical Field

[0001] This application relates to the technical field of Internet of Things (IoT) and industrial safety production monitoring, and in particular to a method and system for intelligent inspection and early warning of safety production that integrates IoT data. Background Technology

[0002] Currently, various IoT monitoring terminals, such as vibration sensors, temperature sensors, gas detectors, and image acquisition equipment, are widely deployed in industrial production sites to continuously monitor the operating status of equipment and the working environment. Periodic inspections are carried out by inspection personnel or inspection equipment to identify potential hazards and anomalies. As the scale of production facilities expands and the complexity of operating conditions increases, on-site monitoring data exhibits characteristics such as multi-source heterogeneity, high timeliness requirements, and diverse anomaly types. Traditional inspection methods that rely on manual experience or a single data source are difficult to meet the needs of high-frequency, accurate, and traceable safety management.

[0003] Existing safety inspection and early warning technology solutions can analyze single-type monitoring data or trigger alarms based on fixed thresholds, but they usually lack the ability to integrate and process multi-source IoT monitoring data in a unified manner. It is difficult to correlate and analyze multi-modal data such as vibration, temperature and environmental conditions under a unified reference, resulting in limited accuracy of anomaly identification results and difficulty in accurately reflecting the overall operating status of equipment.

[0004] The existing technical solutions mentioned above have the following drawbacks: the existing security inspection and early warning methods are difficult to effectively integrate and uniformly assess the status of multi-source IoT monitoring data, resulting in insufficient accuracy in anomaly identification, and therefore there is room for improvement. Summary of the Invention

[0005] To improve the accuracy of safety production inspections, this application provides a method and system for intelligent safety production inspection and early warning that integrates Internet of Things (IoT) data.

[0006] The above-mentioned objective of this application is achieved through the following technical solution:

[0007] A method for intelligent inspection and early warning of safety production that integrates IoT data, comprising:

[0008] Acquire multi-source IoT monitoring data, including equipment operation vibration data, temperature data, and environmental status data;

[0009] The multi-source IoT monitoring data is preprocessed at the edge to obtain preprocessed monitoring data;

[0010] Based on the preprocessed monitoring data, multimodal state features are extracted, and a set of equipment operating state features is constructed. The set of equipment operating state features is then spatiotemporally aligned to obtain fused feature data under a unified reference coordinate.

[0011] The fused feature data is input into a pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation result.

[0012] Based on the equipment operation deviation results, a graded early warning determination is performed to obtain the early warning level, and the corresponding early warning handling strategy is triggered according to the early warning level.

[0013] By adopting the above technical solutions and acquiring multi-source IoT monitoring data, it is possible to comprehensively collect data on equipment operation vibration, temperature, and environmental conditions, thereby forming a complete monitoring foundation covering equipment operation and environmental factors. By performing edge-side preprocessing on the multi-source IoT monitoring data, noise interference can be reduced and data quality can be improved, thereby enhancing the reliability of subsequent analysis. By extracting multimodal state features and performing spatiotemporal alignment processing, heterogeneous data can be unified into the same reference system, thereby enhancing the correlation and comparability between multi-source information. By inputting fused feature data into an anomaly identification model and executing hierarchical early warning judgment, the equipment operating status can be intelligently assessed and targeted handling strategies can be triggered, thereby improving the accuracy of anomaly identification and enhancing the ability to respond to safety risks.

[0014] In one example, this application can be further configured such that: the acquisition of multi-source IoT monitoring data specifically includes:

[0015] Vibration data and temperature data of the equipment are collected according to a preset sampling frequency. The temperature data includes equipment operating temperature data and ambient temperature data.

[0016] According to the preset inspection path and preset scanning angle, environmental status data of the preset inspection area is obtained. The environmental status data includes multispectral image data and environmental gas data.

[0017] The device vibration data, temperature data, and environmental status data are uniformly packaged and processed to form the multi-source IoT monitoring data.

[0018] By adopting the above technical solution, and by collecting equipment vibration data and temperature data at a preset sampling frequency and obtaining environmental status data of the inspection area, it is possible to ensure that equipment status information and environmental information are acquired synchronously, thereby improving the integrity and real-time performance of monitoring data. By uniformly encapsulating and processing vibration, temperature and environmental status data, it is possible to achieve unified format and structured expression of multi-source data, thereby providing a standardized data foundation for subsequent fusion analysis and model input.

[0019] In one example, this application can be further configured such that, prior to acquiring multi-source IoT monitoring data, the intelligent inspection and early warning method for safe production that integrates IoT data further includes:

[0020] Acquire historical inspection data and historical operation records of the equipment, and determine the risk status information of the equipment based on the historical inspection data and the historical operation records of the equipment. The risk status information includes equipment risk level information and risk area distribution information.

[0021] Based on the risk status information, priority inspection areas are determined, and the initial inspection path is adjusted based on the priority inspection areas to obtain the preset inspection path. The initial inspection path is a pre-set default inspection path.

[0022] The initial sampling frequency of the corresponding monitoring node is adaptively adjusted based on the risk status information to obtain the preset sampling frequency, wherein the initial sampling frequency is a pre-set default sampling frequency.

[0023] By adopting the above technical solutions, and by determining the risk status information of equipment based on historical inspection data and equipment operation records, high-risk equipment and areas can be identified in advance, thereby achieving a reasonable allocation of inspection resources. By adjusting the inspection path and sampling frequency according to the risk status information, the inspection strategy can be dynamically matched with the risk level of the equipment, thereby improving inspection efficiency and enhancing the monitoring capability of key risk points.

[0024] In one example, this application can be further configured such that: the edge-side preprocessing of the multi-source IoT monitoring data to obtain preprocessed monitoring data specifically includes:

[0025] The vibration data of the equipment operation is filtered to obtain preprocessed vibration data of the equipment operation.

[0026] Thermal emissivity compensation is performed on the temperature data to obtain preprocessed temperature data;

[0027] The environmental status data is subjected to image quality enhancement and gas data calibration processing to obtain environmental monitoring data;

[0028] The preprocessed equipment operation vibration data, the preprocessed temperature data, and the environmental monitoring data are subjected to timestamp unification, data format conversion, and structured encapsulation to obtain the preprocessed monitoring data.

[0029] By adopting the above technical solutions, filtering vibration data can remove interference components from equipment operation signals, thereby improving the accuracy of vibration characteristics; thermal emissivity compensation of temperature data can reduce measurement errors and improve the authenticity of temperature information, thereby enhancing the reliability of equipment thermal state analysis; image enhancement and gas data calibration of environmental state data can improve the clarity and accuracy of environmental monitoring data, thereby enhancing the ability to identify environmental risks; and unified timestamps and structured data encapsulation can ensure the consistency of multi-source data in the time dimension, thereby improving the effect of subsequent spatiotemporal fusion processing.

[0030] In one example, this application can be further configured such that: the filtering process of the equipment operating vibration data to obtain preprocessed equipment operating vibration data specifically includes:

[0031] Obtain the device type, determine the characteristic frequency band of the vibration signal based on the device type, and dynamically determine the filter cutoff frequency based on the characteristic frequency band;

[0032] The vibration data of the equipment operation is filtered according to the filter cutoff frequency to obtain the filtered vibration data.

[0033] The filtered vibration data is then normalized to obtain normalized vibration data.

[0034] The normalized vibration data is segmented according to a preset vibration feature extraction window to obtain the preprocessed equipment operation vibration data.

[0035] By adopting the above technical solutions, and by determining the characteristic frequency band of the vibration signal based on the equipment type and dynamically setting the filter cutoff frequency, the signal processing strategy can be adapted to the operating characteristics of different equipment, thereby improving the effective information retention rate of the vibration signal. By normalizing the filtered vibration data, the influence of amplitude differences under different equipment or operating conditions can be eliminated, thereby improving feature comparability. By obtaining preprocessed vibration data through segmented processing, a stable feature extraction input sequence can be formed, thereby improving the input stability of the subsequent anomaly recognition model.

[0036] In one example, this application can be further configured as follows: The step of extracting multimodal state features based on the preprocessed monitoring data, constructing a device operating state feature set, and performing spatiotemporal alignment processing on the device operating state feature set to obtain fused feature data under a unified reference coordinate system specifically includes:

[0037] The frequency domain characteristic parameters of the vibration signal are calculated based on the preprocessed equipment operation vibration data to obtain the equipment operation status characteristics;

[0038] Based on the preprocessed temperature data, the temperature distribution characteristics of the equipment operating area are constructed to obtain the thermal state characteristics of the equipment.

[0039] Image region features and gas concentration distribution features are extracted from the environmental monitoring data to obtain the equipment environmental status features;

[0040] The equipment operating status characteristics, the equipment thermal status characteristics, and the equipment environmental status characteristics are aggregated according to the equipment identifier to construct a set of equipment operating status characteristics.

[0041] The device operating status feature set is spatiotemporally aligned to obtain fused feature data under a unified reference coordinate system.

[0042] By adopting the above technical solutions, the mechanical state characteristics of the equipment can be reflected by calculating the frequency domain characteristic parameters of the vibration data, thereby improving the fault mode identification capability; by constructing temperature distribution characteristics, the trend of equipment thermal state changes can be reflected, thereby enhancing the ability to identify overheating or abnormal heat distribution; by extracting image and gas concentration characteristics, the state of the surrounding environment of the equipment can be comprehensively reflected, thereby realizing the collaborative risk assessment of equipment and environment; by aggregating by equipment identification and performing spatiotemporal alignment processing, fused feature data under a unified reference can be formed, thereby improving the overall judgment accuracy of the anomaly identification model.

[0043] In one example, this application can be further configured such that the intelligent inspection and early warning method for safe production that integrates IoT data also includes:

[0044] Acquire historical equipment operation monitoring data, and perform feature extraction processing on the historical equipment operation monitoring data to form a training sample set;

[0045] The training sample set is labeled with status based on historical equipment operation records and historical fault records to obtain the training dataset.

[0046] The anomaly detection model is iteratively trained using the training dataset to obtain the pre-built anomaly detection model, which includes a feature input layer, a state modeling layer, and a result output layer.

[0047] During the training of the anomaly detection model, the model parameters are updated based on the loss function using a gradient optimization algorithm, which includes the Adam optimization algorithm.

[0048] By adopting the above technical solutions, and by acquiring historical equipment operation monitoring data and forming a training sample set, a model training foundation that reflects the actual operating status of the equipment can be constructed, thereby improving the model's generalization ability. By combining historical operation records and fault records for status labeling, the training data can have clear health labels, thereby improving the model's accuracy in identifying abnormal patterns. By iteratively training the anomaly identification model and using gradient optimization algorithms to update parameters, the model performance can be continuously optimized, thereby improving the accuracy and stability of anomaly identification.

[0049] In one example, this application can be further configured as follows: the step of performing a graded early warning determination based on the equipment operation deviation result to obtain an early warning level, and triggering a corresponding early warning handling strategy based on the early warning level, specifically includes:

[0050] The equipment operating status level is determined by comparing the equipment operating deviation result with a preset health threshold range.

[0051] Map the device's operating status level to the corresponding warning level;

[0052] According to the warning level, a corresponding warning handling strategy is matched from the preset warning strategy library and executed. The warning handling strategy includes at least one of on-site prompt warning, remote notification warning and safety linkage control.

[0053] By adopting the above technical solutions, the equipment status can be quantitatively assessed by comparing the equipment operation deviation results with the health threshold range, thereby improving the objectivity of early warning judgment; by mapping the equipment status level to the early warning level, the analysis results can be transformed into a unified risk level, thereby improving the understandability of early warning information; by matching and executing the disposal strategy from the preset early warning strategy library, the early warning results can directly correspond to the actual disposal measures, thereby improving the timeliness and effectiveness of safety response.

[0054] In one example, this application can be further configured such that the intelligent inspection and early warning method for safe production that integrates IoT data also includes:

[0055] Based on the equipment operation deviation results and early warning levels, generate equipment risk distribution information;

[0056] The equipment risk distribution information is mapped to the spatial layout of the inspection area to construct an equipment risk distribution map, which is then displayed through a visual interface.

[0057] By adopting the above technical solutions, and generating equipment risk distribution information based on equipment operation deviation results and warning levels, the spatial distribution of equipment risks can be intuitively reflected, thereby improving the overall effectiveness of risk identification. By mapping risk information to inspection areas and constructing risk distribution maps for visualization, managers can quickly locate high-risk areas, thereby improving inspection decision-making efficiency and enhancing the visualization level of safety management.

[0058] The second objective of this invention is achieved through the following technical solution:

[0059] A smart safety inspection and early warning system integrating IoT data, comprising:

[0060] The monitoring data acquisition module is used to acquire multi-source IoT monitoring data, which includes equipment operation vibration data, temperature data, and environmental status data.

[0061] An edge preprocessing module is used to perform edge-side preprocessing on the multi-source IoT monitoring data to obtain preprocessed monitoring data;

[0062] The multimodal feature construction module is used to extract multimodal state features based on the preprocessed monitoring data, construct a set of equipment operating state features, and perform spatiotemporal alignment processing on the set of equipment operating state features to obtain fused feature data under a unified reference coordinate.

[0063] An anomaly assessment module is used to input the fused feature data into a pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation result.

[0064] The early warning determination module is used to perform graded early warning determination based on the equipment operation deviation results, obtain the early warning level, and trigger the corresponding early warning handling strategy according to the early warning level.

[0065] By adopting the above technical solutions and acquiring multi-source IoT monitoring data, it is possible to comprehensively collect data on equipment operation vibration, temperature, and environmental conditions, thereby forming a complete monitoring foundation covering equipment operation and environmental factors. By performing edge-side preprocessing on the multi-source IoT monitoring data, noise interference can be reduced and data quality can be improved, thereby enhancing the reliability of subsequent analysis. By extracting multimodal state features and performing spatiotemporal alignment processing, heterogeneous data can be unified into the same reference system, thereby enhancing the correlation and comparability between multi-source information. By inputting fused feature data into an anomaly identification model and executing hierarchical early warning judgment, the equipment operating status can be intelligently assessed and targeted handling strategies can be triggered, thereby improving the accuracy of anomaly identification and enhancing the ability to respond to safety risks.

[0066] In summary, this application includes the following beneficial technical effects:

[0067] 1. By acquiring multi-source IoT monitoring data, comprehensive collection of equipment operation vibration, temperature, and environmental conditions can be achieved, thus forming a complete monitoring foundation covering equipment operation and environmental factors. By performing edge-side preprocessing on the multi-source IoT monitoring data, noise interference can be reduced and data quality can be improved, thereby enhancing the reliability of subsequent analysis. By extracting multimodal state features and performing spatiotemporal alignment processing, heterogeneous data can be unified into the same reference system, thereby enhancing the correlation and comparability between multi-source information. By inputting fused feature data into an anomaly identification model and executing hierarchical early warning judgment, intelligent assessment of equipment operating status can be performed and targeted handling strategies can be triggered, thereby improving the accuracy of anomaly identification and enhancing the ability to respond to safety risks.

[0068] 2. By determining equipment risk status information based on historical inspection data and equipment operation records, high-risk equipment and areas can be identified in advance, thereby achieving a reasonable allocation of inspection resources; by adjusting inspection paths and sampling frequencies according to risk status information, inspection strategies can be dynamically matched with equipment risk levels, thereby improving inspection efficiency and enhancing the monitoring capabilities of key risk points.

[0069] 3. By generating equipment risk distribution information based on equipment operation deviation results and warning levels, the spatial distribution of equipment risks can be intuitively reflected, thereby improving the overall effectiveness of risk identification. By mapping risk information to inspection areas and constructing risk distribution maps for visualization, managers can quickly locate high-risk areas, thereby improving inspection decision-making efficiency and enhancing the visualization level of safety management. Attached Figure Description

[0070] Figure 1 This is a flowchart of a method for intelligent inspection and early warning of safe production that integrates Internet of Things data in one embodiment of this application;

[0071] Figure 2 This is a principle block diagram of a smart inspection and early warning system for safe production that integrates Internet of Things data in one embodiment of this application. Detailed Implementation

[0072] The present application will be further described in detail below with reference to the accompanying drawings.

[0073] In one embodiment, such as Figure 1 As shown, this application discloses a method for intelligent inspection and early warning of safe production that integrates Internet of Things (IoT) data, specifically including the following steps:

[0074] S10: Acquire multi-source IoT monitoring data, including equipment operation vibration data, temperature data, and environmental status data.

[0075] Specifically, vibration signals during equipment operation are collected by vibration acquisition units deployed on the equipment surface and continuously collected at a preset sampling rate to form a vibration data stream. Infrared temperature measurement units collect the surface temperature of the equipment and simultaneously acquire the ambient temperature around the equipment to form temperature data. Environmental monitoring units set up in the inspection area acquire multispectral image information and gas concentration information to form environmental status data. The vibration acquisition units can use high-bandwidth MEMS vibration sensors to acquire wide-band mechanical vibration signals. The temperature acquisition process can acquire the trend of thermal state changes of the equipment through radiation temperature measurement. The environmental monitoring units can identify abnormal hot spots or corrosion areas on the equipment surface through multispectral imaging. At the same time, gas sensors can detect leaks or abnormal gas concentration changes. For example, in the scenario of conveyor bearing monitoring, the vibration acquisition unit can acquire the bearing vibration spectrum changes to determine the wear trend, the temperature measurement unit can detect the local temperature rise of the bearing to identify insufficient lubrication, and the gas sensor can detect combustible gas leaks to identify environmental risks, thereby realizing multi-dimensional synchronous acquisition of equipment operating status, thermal status, and environmental status.

[0076] S20: Perform edge-side preprocessing on multi-source IoT monitoring data to obtain preprocessed monitoring data.

[0077] Specifically, the acquired vibration signals are filtered to suppress mechanical noise and environmental interference during the acquisition process and improve the signal-to-noise ratio. Emissivity compensation is performed on the temperature data to correct temperature measurement errors caused by the radiation characteristics of different material surfaces. Contrast enhancement and noise suppression are performed on the multispectral image data to improve image clarity and the accuracy of subsequent feature extraction. At the same time, calibration compensation is performed on the gas concentration data to eliminate sensor drift errors. For example, in the pipeline leak monitoring scenario, image enhancement can more clearly identify corrosion spots on the pipeline surface. Gas data calibration can avoid false alarms caused by sensor aging. After completing various data preprocessing, the data are aligned to a unified time reference and standardized in format, and then structured and encapsulated to form preprocessed monitoring data in a unified format for subsequent multimodal analysis.

[0078] S30: Extract multimodal state features based on preprocessed monitoring data, construct a set of equipment operating state features, and perform spatiotemporal alignment processing on the set of equipment operating state features to obtain fused feature data under a unified reference coordinate system.

[0079] Specifically, frequency domain energy distribution, dominant frequency components, and vibration amplitude variation trends are extracted from vibration data to characterize the mechanical operating status of the equipment. Temperature gradient distribution and local temperature rise variation trends are extracted from temperature data to reflect the thermal load status of the equipment. Texture and thermal anomaly features are extracted from multispectral images to identify potential fault signs on the equipment surface. Concentration change rate and diffusion trend are extracted from gas concentration data to identify potential leakage risks. For example, in motor operation monitoring, rotor imbalance can be identified by vibration spectrum changes, coil overload can be identified by temperature rise gradients, and insulation damage can be identified by image anomaly areas. The above features are associated and aggregated according to equipment identification to construct a set of equipment operating status features. Alignment processing is performed using a unified time reference and spatial position relationship to eliminate time differences and spatial offsets in multi-source data acquisition, thereby obtaining fused feature data under a unified reference coordinate to reflect the comprehensive operating status of the equipment.

[0080] S40: Input the fused feature data into the pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation results.

[0081] Specifically, the fused feature data is used as the input feature vector and fed into the anomaly identification model to perform state evaluation calculation. The anomaly identification model evaluates the current equipment state by learning the mapping relationship between historical equipment operating states and failure modes and outputs the deviation result representing the health level of the equipment. The deviation reflects the degree of difference between the current equipment state and the normal operating state. When the deviation increases, it indicates that the equipment operating state has an abnormal trend. For example, in the monitoring of wind turbine operation, when the vibration spectrum energy continues to increase in the abnormal frequency band and the temperature gradient increases synchronously, the model can determine that the equipment has entered the abnormal trend stage, thereby realizing the quantitative assessment of the potential failure risk of the equipment.

[0082] S50: Based on the equipment operation deviation results, perform a graded early warning judgment to obtain the early warning level, and trigger the corresponding early warning handling strategy according to the early warning level.

[0083] Specifically, the deviation results of equipment operation are compared with the preset health range to determine the current status level of the equipment. When the deviation is within the normal range, the monitoring status is maintained. When the deviation enters the slightly abnormal range, a prompt-level warning is triggered to remind maintenance personnel to pay attention to the equipment status. When the deviation further increases and enters the severely abnormal range, a notification-level warning is triggered to send alarm information to the remote terminal and prompt to arrange maintenance. When the deviation reaches the danger range, a linkage control strategy is triggered to execute equipment load reduction, shutdown, or safety isolation operations, thereby realizing a graded response to equipment abnormal risks and reducing the probability of accidents.

[0084] In another embodiment, to further enhance the inspection coverage and multi-dimensional monitoring accuracy, an inspection scheme combining UAV inspection and acoustic emission monitoring can be adopted. The UAV inspection unit uses a carbon fiber frame to reduce weight and increase structural strength. The unit is equipped with flight control and edge computing components based on the Qualcomm Flight_RB5_5G platform for real-time data processing and wireless communication. The navigation section integrates an RTK-GNSS high-precision positioning module and a TOF lidar for centimeter-level spatial positioning and obstacle detection. The power supply uses a high-capacity lithium polymer battery and is equipped with a ground-based fast-charging device to support continuous inspection missions. The acoustic emission monitoring unit uses a high-sensitivity acoustic emission sensor and a high-gain preamplifier to improve the acquisition accuracy of weak structural acoustic signals. Simultaneously, a signal processing method based on time-frequency joint analysis is used to extract features from the acquired signals. The system identifies structural damage or impact events and generates acoustic emission monitoring results. During deployment, it can be implemented in stages according to the layout of on-site equipment. In the hardware deployment stage, fixed monitoring nodes are set up based on the length of the inspection area and the distribution of equipment, and the operating speed and inspection interval of the mobile inspection unit are set to ensure continuous monitoring. Edge computing devices can be installed in a protective control box to adapt to temperature changes in the industrial environment. In the software configuration stage, data processing and storage modules are deployed through a cloud computing platform, and a distributed database strategy is adopted to improve data access efficiency. Simultaneously, mobile terminal applications are developed to achieve real-time early warning information push. In the joint debugging and testing stage, the inspection and early warning capabilities are verified by simulating typical equipment failure states. For example, the vibration and temperature characteristic recognition capabilities can be verified by simulating insufficient bearing lubrication or abnormal structural temperature rise. Overall performance is evaluated based on indicators such as response time and false alarm rate to confirm the deployment effect.

[0085] In one embodiment, step S10, namely acquiring multi-source IoT monitoring data, specifically includes:

[0086] S11: Collect equipment vibration data and temperature data according to the preset sampling frequency. The temperature data includes equipment operating temperature data and ambient temperature data.

[0087] Specifically, the sampling frequency is set according to the equipment type and monitoring target, and the vibration acquisition unit is driven to periodically acquire the vibration signal of the equipment operation. At the same time, the temperature measurement unit is driven to synchronously acquire the surface temperature of the equipment and the ambient temperature of the equipment to form temperature data. For example, in the monitoring scenario of high-speed rotating equipment, the vibration sampling frequency can be increased to capture the early failure characteristics of the bearing. In the monitoring of low-speed conveying equipment, a lower sampling frequency can be used to reduce data redundancy. By acquiring the equipment temperature and the ambient temperature at the same time, the ambient temperature fluctuation can avoid misleading the judgment of the thermal state of the equipment, thereby ensuring that the collected data can truly reflect the operating status of the equipment.

[0088] S12: Obtain environmental status data of the preset inspection area according to the preset inspection path and preset scanning angle. The environmental status data includes multispectral image data and environmental gas data.

[0089] Specifically, the inspection device is driven to move to each monitoring node according to the preset inspection path and to collect images of the inspection area at the preset scanning angle to obtain multispectral image data. At the same time, gas concentration information is collected at the inspection node to form environmental gas data. For example, in the inspection scenario of chemical storage tank, images can be collected along the circumferential path of the storage tank to identify corrosion spots or leakage traces, and gas concentration can be collected near the valve to determine whether there is any abnormality in volatile gas. By controlling the path and angle, it can be ensured that the inspection area is completely covered and the detection blind spots caused by obstruction can be reduced, thereby achieving a comprehensive perception of the environmental status.

[0090] S13: Unify and encapsulate equipment vibration data, temperature data, and environmental status data to form multi-source IoT monitoring data.

[0091] Specifically, the collected vibration data, temperature data, and environmental status data are timestamped and associated with the device identifier. At the same time, different types of data are converted into a unified data format and written into a unified data structure. For example, vibration spectrum data, temperature time series data, and multispectral image data are converted into structured fields and appended with the acquisition time and device number. Through unified encapsulation, it can be ensured that the data source and time relationship can be accurately identified in subsequent processing, thereby forming multi-source IoT monitoring data that can be used for subsequent analysis and processing.

[0092] In one embodiment, before step S11, i.e. before acquiring multi-source IoT monitoring data, the intelligent inspection and early warning method for safe production that integrates IoT data further includes:

[0093] S1101: Obtain historical inspection data and historical operation records of the equipment, and determine the risk status information of the equipment based on the historical inspection data and historical operation records. The risk status information includes the equipment risk level information and the risk area distribution information.

[0094] Specifically, the system reads the inspection results recorded during historical inspections and the operation logs during equipment operation. It then performs statistical analysis on the abnormal records, number of faults, and maintenance frequency to assess the stability of equipment operation. Simultaneously, it combines the location of abnormalities with spatial correlation analysis of the equipment's location to form a risk area distribution. For example, if a pump station has multiple vibration abnormalities and frequent maintenance records in past inspections, its risk level can be determined to be high, and its location can be marked as a key risk area. By comprehensively analyzing historical data, risk status information reflecting the true risk level of the equipment can be generated.

[0095] S1102: Determine the priority inspection area based on the risk status information, and adjust the initial inspection path based on the priority inspection area to obtain the preset inspection path. The initial inspection path is the pre-set default inspection path.

[0096] Specifically, based on the equipment risk level and risk area distribution information, the inspection areas are prioritized and key areas that need to be inspected first are identified. Then, based on the initial inspection path, the order of the path nodes is rearranged and the number of inspections of high-risk areas is increased to form a new inspection path. For example, when an area is identified as a high-risk area, its inspection nodes can be moved to the beginning of the path and the number of times it is passed can be increased to ensure priority inspection. By dynamically adjusting the path, the utilization efficiency of inspection resources can be improved and high-risk equipment can be monitored first.

[0097] S1103: Adaptively adjust the initial sampling frequency of the corresponding monitoring node based on the risk status information to obtain the preset sampling frequency. The initial sampling frequency is the preset default sampling frequency.

[0098] Specifically, the sampling cycle of each monitoring node is set differently according to the risk level of the equipment, and the data acquisition frequency of high-risk nodes is increased while the sampling frequency of low-risk nodes is appropriately reduced to reduce redundant data. For example, when a device is determined to be in a high-risk state, its vibration data sampling cycle can be shortened to capture abnormal changes more promptly, while a lower sampling frequency can be maintained for devices that have been operating stably for a long time. By adaptively adjusting the sampling frequency, the data processing burden can be reduced and the overall monitoring efficiency can be improved while ensuring monitoring accuracy.

[0099] In one embodiment, step S20, which involves edge-side preprocessing of the multi-source IoT monitoring data to obtain preprocessed monitoring data, specifically includes:

[0100] S21: Filter the equipment operation vibration data to obtain preprocessed equipment operation vibration data.

[0101] Specifically, the original vibration signal collected by the equipment vibration sensor is read and its spectrum is analyzed to identify the noise component and effective vibration component in the signal. Then, based on the preset filtering model, high-frequency random interference and low-frequency drift signal are suppressed and effective frequency band information reflecting the equipment operating status is retained. For example, when the vibration signal contains environmental impact noise, the main mechanical vibration frequency range can be retained by bandpass filtering to obtain stable and reliable vibration data for subsequent analysis.

[0102] S22: Perform thermal emissivity compensation on the temperature data to obtain preprocessed temperature data.

[0103] Specifically, the raw temperature value measured by the temperature acquisition device is obtained and the measurement results are corrected by combining the material properties of the equipment surface and the influence of environmental thermal radiation to eliminate measurement deviations caused by differences in surface reflection or radiation. At the same time, the measured value is dynamically compensated according to the trend of environmental temperature change. For example, when the metal casing equipment is working in a high-temperature environment, the measured value can be corrected according to its radiation characteristics to obtain data that is closer to the actual operating temperature, thereby improving the accuracy of temperature monitoring.

[0104] S23: Perform image quality enhancement and gas data calibration on the environmental status data to obtain environmental monitoring data.

[0105] Specifically, the acquired image data undergoes brightness equalization and noise suppression processing to improve image clarity and enhance the recognizability of key area features. At the same time, the acquired gas concentration data is calibrated and corrected according to the calibration curve of the gas sensor to eliminate drift errors and environmental interference. For example, when the contrast of the inspection image decreases due to insufficient lighting, the equipment outline information can be highlighted through enhancement processing. When the gas detection value is affected by temperature and humidity, its concentration reading can be corrected through the calibration curve to obtain more reliable environmental monitoring data.

[0106] S24: The preprocessed equipment operation vibration data, preprocessed temperature data, and environmental monitoring data are subjected to timestamp unification, data format conversion, and structured encapsulation to obtain preprocessed monitoring data.

[0107] Specifically, based on the acquisition time information of various sensor data, data from different sources are time-aligned and uniformly converted into a standard time series format. At the same time, different types of data are encapsulated according to a unified field structure to form data records that can be directly used for subsequent analysis. For example, when vibration data is sampled at the millisecond level and temperature data is sampled at the second level, they can be mapped to a unified time axis through time alignment and combined and stored in a structured form to ensure that there is a consistent time reference relationship between multi-source data, thereby improving the accuracy of data fusion processing.

[0108] In one embodiment, step S21, which involves filtering the equipment operating vibration data to obtain preprocessed equipment operating vibration data, specifically includes:

[0109] S211: Obtain the device type, determine the characteristic frequency band of the vibration signal based on the device type, and dynamically determine the filter cutoff frequency based on the characteristic frequency band.

[0110] Specifically, the system reads the basic information of the equipment and parses the equipment category identifier to match the corresponding vibration characteristic model. It retrieves the typical vibration frequency range of the equipment of this category from the equipment operation database and determines the effective frequency band in the signal accordingly. At the same time, it automatically calculates the appropriate filter cutoff frequency based on the effective frequency band to avoid interference from invalid frequencies. For example, when the equipment is a rotating motor, the filter interval can be determined based on the main frequency range of mechanical vibration corresponding to its rotation speed, so that the subsequent processing focuses on the vibration signal that reflects the actual operating state.

[0111] S212: Filter the equipment operation vibration data according to the filter cutoff frequency to obtain the filtered vibration data.

[0112] Specifically, based on the determined filter cutoff frequency, digital filtering is performed on the acquired vibration time series signal to suppress irrelevant frequency components and retain key frequency information reflecting the equipment's operating status. At the same time, abnormal and sudden signals are smoothed to reduce the impact of sampling noise. For example, when there is environmental impact noise in the vibration data, filtering can weaken the instantaneous interference peak and obtain more stable vibration waveform data for subsequent analysis.

[0113] S213: Perform amplitude normalization on the filtered vibration data to obtain normalized vibration data.

[0114] Specifically, amplitude scaling is performed on the filtered vibration signal to map the vibration intensity of different sampling periods and different devices to a unified numerical range and eliminate the influence of sensor range differences. At the same time, the signal change trend is kept unchanged to ensure the consistency of subsequent feature extraction. For example, when the vibration value range of different devices differs greatly, normalization can bring them to the same analysis scale, thereby improving the stability of subsequent model processing.

[0115] S214: The normalized vibration data is segmented according to the preset vibration feature extraction window to obtain the preprocessed equipment operation vibration data.

[0116] Specifically, the continuous vibration data is processed by sliding segmentation according to the preset time window length to form multiple signal segments with fixed time spans. The corresponding time identifier information of each segment is retained for subsequent feature extraction and state analysis. At the same time, the data continuity is ensured during the window sliding process to avoid information loss. For example, the data can be segmented in units of several seconds to form stable data blocks, thereby providing a unified input format for subsequent vibration feature calculation.

[0117] In one embodiment, step S30 involves extracting multimodal state features based on the preprocessed monitoring data, constructing a device operating state feature set, and performing spatiotemporal alignment processing on the device operating state feature set to obtain fused feature data under a unified reference coordinate system. Specifically, this includes:

[0118] S31: Calculate the frequency domain characteristic parameters of the vibration signal based on the preprocessed equipment operation vibration data to obtain the equipment operation status characteristics.

[0119] Specifically, frequency domain analysis is performed on the preprocessed vibration time series to obtain the spectral distribution information of the signal and calculate key frequency domain indicators that reflect the stability of equipment operation. At the same time, vibration energy distribution and dominant frequency characteristics are extracted to characterize the mechanical operating state of the equipment. For example, when there is bearing wear in the equipment, it can be manifested as abnormal energy enhancement in a specific frequency band, so that the extracted frequency domain characteristics can reflect the potential mechanical failure trend.

[0120] S32: Construct the temperature distribution characteristics of the equipment operating area based on the preprocessed temperature data to obtain the thermal state characteristics of the equipment.

[0121] Specifically, the temperature data after compensation processing is spatially mapped according to the collection location, and a temperature field distribution model is constructed to describe the thermal state changes of various parts of the equipment. At the same time, the temperature gradient and local thermal anomaly index are calculated to reflect the thermal load during equipment operation. For example, when a local area of ​​the equipment experiences heat dissipation anomalies, its temperature distribution will show uneven changes, so that the constructed temperature features can characterize the thermal risk status of the equipment.

[0122] S33: Extract image region features and gas concentration distribution features from environmental monitoring data to obtain equipment environmental status features.

[0123] Specifically, regional feature extraction operations are performed on environmental image data to identify the operating environment status around the equipment, and spatial analysis is performed on gas concentration data to obtain information on environmental change trends, thereby forming an environmental feature description that can reflect the external safety status of the equipment. For example, when there is smoke diffusion or abnormal gas accumulation around the equipment, its image texture features and gas concentration changes will present an abnormal pattern, so that the extracted environmental features can characterize potential safety hazards.

[0124] S34: Aggregate the equipment operating status characteristics, equipment thermal status characteristics, and equipment environmental status characteristics according to the equipment identifier to construct the equipment operating status characteristic set.

[0125] Specifically, based on the unique identifier of the device, multimodal features from different sensor sources are associated and bound and uniformly organized into a comprehensive feature description corresponding to the same device. At the same time, various features are integrated in chronological order to form a continuous status record. For example, when the same device experiences both abnormal vibration and temperature rise at the same time, feature aggregation can form a complete expression of the operating status, thereby avoiding misjudgment caused by a single data source.

[0126] S35: Perform spatiotemporal alignment processing on the equipment operating status feature set to obtain fused feature data under a unified reference coordinate system.

[0127] Specifically, the aggregated multimodal features are uniformly calibrated based on the acquisition timestamp and spatial location, and the offset effect caused by the difference in sampling time and acquisition location of different data sources is eliminated. At the same time, a unified reference coordinate is constructed to ensure that various features correspond under the same analysis benchmark. For example, when the vibration data acquisition frequency is higher than the image acquisition frequency, time alignment processing can make various data correspond within the same time window, thereby forming fused feature data that can be directly used for subsequent state assessment.

[0128] In one embodiment, the intelligent inspection and early warning method for safe production that integrates IoT data further includes:

[0129] S401: Acquire historical equipment operation monitoring data, and perform feature extraction processing on the historical equipment operation monitoring data to form a training sample set.

[0130] Specifically, historical equipment operation monitoring data is read and parsed according to a unified data format to obtain equipment operation behavior records over a continuous period of time. At the same time, feature information that reflects the equipment's operating status is extracted based on the data change patterns, and a sample structure is constructed for subsequent model learning. For example, when the vibration amplitude of the equipment gradually increases during long-term operation, its characteristic change trend can be extracted and used to form training samples that reflect potential anomalies.

[0131] S402: The training sample set is labeled with its status based on historical equipment operation records and historical fault records to obtain the training dataset.

[0132] Specifically, the training samples are correlated with the equipment operation logs and the historical fault occurrence times to determine the corresponding operating status of the samples and assign label information to each sample. At the same time, normal operation samples and abnormal operation samples are distinguished to form a data structure with discriminative ability. For example, when a device has a downtime record within a certain period of time, the corresponding sample of that period of time can be marked as an abnormal state, so that the training data has a basis for supervised learning.

[0133] S403: Iteratively train the anomaly recognition model using the training dataset to obtain a pre-built anomaly recognition model, which includes a feature input layer, a state modeling layer, and a result output layer.

[0134] Specifically, the training dataset is input into the model structure, and the internal parameters of the model are gradually adjusted through multiple rounds of iterative calculations to learn the mapping relationship between the device's operating state and abnormal modes. At the same time, multimodal feature vectors are received at the input layer and feature representation learning is performed at the state modeling layer to form a state representation. Then, the abnormal probability results are generated through the output layer. For example, when there is a feature change pattern before the device malfunction in the training samples, the model can gradually learn the pattern during the iteration process to improve the subsequent recognition ability.

[0135] S404: During the training of the anomaly detection model, the model parameters are updated based on the loss function using a gradient optimization algorithm, which includes the Adam optimization algorithm.

[0136] Specifically, during the model training iteration process, the error value between the predicted result and the true label is calculated, and parameter update operations are performed according to the direction of error change to gradually reduce the model prediction bias. At the same time, the adaptive learning rate adjustment strategy is combined to improve the training convergence speed and stability. For example, when the model has large error fluctuations in the early training stage, the parameter adjustment can be made more stable through adaptive gradient update, thereby avoiding oscillations in the training process and improving the model's generalization ability.

[0137] In one embodiment, step S50 involves performing a graded early warning determination based on the equipment operation deviation result to obtain the early warning level, and triggering the corresponding early warning handling strategy according to the early warning level, specifically including:

[0138] S51: Determine the equipment operating status level by comparing the equipment operation deviation results with the preset health threshold range.

[0139] Specifically, numerical analysis is performed on the deviation results of equipment operation, and the results are compared with the preset health threshold range to determine the risk level of the current operating status of the equipment. At the same time, the status level range is divided according to the degree of deviation to reflect the trend of equipment health changes. For example, when the deviation value is within the normal threshold range, it is judged as a healthy state, while when the deviation continues to approach the critical threshold, it is judged as a warning state, thus forming a status level result that can be used for risk assessment.

[0140] S52: Map the equipment operating status level to the corresponding warning level.

[0141] Specifically, a mapping relationship between the status and the warning level is established based on the equipment's operating status level, and the corresponding warning level is automatically generated according to the mapping rules to achieve a unified conversion between risk level and handling level. At the same time, the mapping mechanism enables different equipment types to be comparable under a unified warning framework. For example, when the equipment status level is determined to be a high-risk status, it can be directly mapped to a high-level warning level, thereby triggering stricter monitoring and response strategies.

[0142] S53: Match the corresponding early warning handling strategy from the preset early warning strategy library according to the early warning level, and execute the early warning handling strategy. The early warning handling strategy includes at least one of on-site prompt early warning, remote notification early warning and safety linkage control.

[0143] Specifically, based on the warning level, the system retrieves a corresponding response plan from a pre-set set of strategies that matches the current risk level and generates an execution command to complete the warning response process. At the same time, it selects different intervention intensities based on different warning levels to achieve risk grading control. For example, when the equipment is under a medium-level warning, it triggers on-site prompts to remind inspection personnel to pay attention to the equipment status, while when the equipment is under a high-level warning, it can simultaneously generate remote notifications and link with the control of relevant safety equipment to reduce the probability of accidents.

[0144] In one embodiment, the intelligent inspection and early warning method for safe production that integrates IoT data further includes:

[0145] S60: Generate equipment risk distribution information based on equipment operation deviation results and early warning levels.

[0146] Specifically, the risk status of each piece of equipment is quantified based on the deviation value of the equipment operation and the corresponding warning level, and risk description data containing equipment identification, risk level and risk weight is generated to form equipment risk distribution information. At the same time, the comparability between different equipment is made by unifying the risk expression method. For example, different risk weights can be assigned to equipment according to the deviation size, and high-risk equipment identification can be generated by combining the warning level, thereby forming a data set that can reflect the risk status distribution of the inspection area.

[0147] S70: Map the equipment risk distribution information to the spatial layout of the inspection area, construct an equipment risk distribution map, and display the equipment risk distribution map through a visual interface.

[0148] Specifically, based on the spatial coordinates of the equipment in the inspection area, the risk distribution information of the equipment is mapped to the corresponding area and a spatialized risk expression result is generated to construct an equipment risk distribution map. At the same time, the risk status is visualized in a graphical way. For example, high-risk equipment can be displayed in red in the inspection area layout map and medium-risk equipment can be displayed in orange, so that inspection personnel can intuitively identify the risk concentration area and assist in determining the inspection priority order.

[0149] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0150] In one embodiment, a smart inspection and early warning system for safe production that integrates IoT data is provided. This smart inspection and early warning system for safe production that integrates IoT data corresponds one-to-one with the smart inspection and early warning method for safe production that integrates IoT data in the above embodiments. Figure 2 As shown, this intelligent safety inspection and early warning system integrating IoT data includes a monitoring data acquisition module, an edge preprocessing module, a multimodal feature construction module, an anomaly assessment module, and an early warning determination module. Detailed descriptions of each functional module are as follows:

[0151] The monitoring data acquisition module is used to acquire multi-source IoT monitoring data, which includes equipment operation vibration data, temperature data, and environmental status data.

[0152] The edge preprocessing module is used to perform edge-side preprocessing on multi-source IoT monitoring data to obtain preprocessed monitoring data;

[0153] The multimodal feature construction module is used to extract multimodal state features based on preprocessed monitoring data, construct a set of equipment operating state features, and perform spatiotemporal alignment processing on the set of equipment operating state features to obtain fused feature data under a unified reference coordinate.

[0154] The anomaly assessment module is used to input fused feature data into a pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation results.

[0155] The early warning determination module is used to perform graded early warning determination based on the equipment operation deviation results, obtain the early warning level, and trigger the corresponding early warning handling strategy according to the early warning level.

[0156] Optionally, the monitoring data acquisition module includes:

[0157] The vibration and temperature acquisition submodule is used to acquire equipment vibration data and temperature data according to a preset sampling frequency. The temperature data includes equipment operating temperature data and ambient temperature data.

[0158] The environmental inspection and data acquisition submodule is used to acquire environmental status data of a preset inspection area according to a preset inspection path and preset scanning angle. The environmental status data includes multispectral image data and environmental gas data.

[0159] The data encapsulation submodule is used to encapsulate and process equipment vibration data, temperature data, and environmental status data in a unified manner to form multi-source IoT monitoring data.

[0160] Optionally, the intelligent inspection and early warning system for safe production that integrates IoT data also includes:

[0161] The historical risk analysis module is used to acquire historical inspection data and historical operation records of the equipment, and to determine the risk status information of the equipment based on the historical inspection data and historical operation records. The risk status information includes the equipment risk level information and the risk area distribution information.

[0162] The inspection path optimization module is used to determine the inspection priority area based on the risk status information, and adjust the initial inspection path based on the inspection priority area to obtain the preset inspection path. The initial inspection path is the pre-set default inspection path.

[0163] The sampling frequency adjustment module is used to adaptively adjust the initial sampling frequency of the corresponding monitoring node according to the risk status information to obtain the preset sampling frequency. The initial sampling frequency is the preset default sampling frequency.

[0164] Optional, the edge preprocessing module includes:

[0165] The vibration signal preprocessing submodule is used to filter the equipment operation vibration data to obtain preprocessed equipment operation vibration data.

[0166] The temperature compensation processing submodule is used to perform thermal emissivity compensation on the temperature data to obtain preprocessed temperature data.

[0167] The environmental data correction submodule is used to perform image quality enhancement and gas data calibration on environmental status data to obtain environmental monitoring data.

[0168] The data standardization and encapsulation submodule is used to perform timestamp unification, data format conversion, and structured encapsulation on the pre-processed equipment operation vibration data, pre-processed temperature data, and environmental monitoring data to obtain pre-processed monitoring data.

[0169] Optionally, the vibration signal preprocessing submodule includes:

[0170] The equipment frequency band identification unit is used to obtain the equipment type, determine the characteristic frequency band of the vibration signal based on the equipment type, and dynamically determine the filter cutoff frequency based on the characteristic frequency band.

[0171] The vibration filtering processing unit is used to filter the equipment operating vibration data according to the filtering cutoff frequency to obtain the filtered vibration data.

[0172] The vibration normalization unit is used to normalize the amplitude of the filtered vibration data to obtain normalized vibration data.

[0173] The vibration segmentation unit is used to segment the normalized vibration data according to the preset vibration feature extraction window to obtain preprocessed equipment operation vibration data.

[0174] Optional, multimodal feature building modules include:

[0175] The vibration feature extraction submodule is used to calculate the frequency domain feature parameters of the vibration signal based on the preprocessed equipment operation vibration data, and obtain the equipment operation status features.

[0176] The temperature feature modeling submodule is used to construct the temperature distribution characteristics of the equipment operating area based on the preprocessed temperature data, thereby obtaining the thermal state characteristics of the equipment.

[0177] The environmental feature extraction submodule is used to extract image region features and gas concentration distribution features from environmental monitoring data to obtain the environmental status features of the equipment.

[0178] The feature aggregation submodule is used to aggregate equipment operating status features, equipment thermal status features, and equipment environmental status features according to the equipment identifier to construct a set of equipment operating status features.

[0179] The spatiotemporal alignment submodule is used to perform spatiotemporal alignment processing on the device operating status feature set to obtain fused feature data under a unified reference coordinate system.

[0180] Optionally, the intelligent inspection and early warning system for safe production that integrates IoT data also includes:

[0181] The training sample construction module is used to acquire historical equipment operation monitoring data and perform feature extraction processing on the historical equipment operation monitoring data to form a training sample set;

[0182] The status labeling module is used to label the training sample set with status based on historical equipment operation records and historical fault records to obtain the training dataset.

[0183] The model training module is used to iteratively train the anomaly recognition model using the training dataset to obtain a pre-built anomaly recognition model, which includes a feature input layer, a state modeling layer, and a result output layer.

[0184] The parameter optimization module is used to update the model parameters based on the loss function and the gradient optimization algorithm during the training of the anomaly detection model. The gradient optimization algorithm includes the Adam optimization algorithm.

[0185] Optionally, the early warning determination module includes:

[0186] The health status comparison submodule is used to compare the equipment operation deviation results with a preset health threshold range to determine the equipment operation status level.

[0187] The early warning mapping submodule is used to map the device's operating status level to the corresponding early warning level;

[0188] The strategy matching and execution submodule is used to match the corresponding early warning handling strategy from the preset early warning strategy library according to the early warning level, and execute the early warning handling strategy. The early warning handling strategy includes at least one of on-site prompt early warning, remote notification early warning and safety linkage control.

[0189] Optionally, the intelligent inspection and early warning system for safe production that integrates IoT data also includes:

[0190] The risk information generation module is used to generate equipment risk distribution information based on equipment operation deviation results and warning levels.

[0191] The risk map construction module is used to map equipment risk distribution information to the spatial layout of the inspection area, construct an equipment risk distribution map, and display the equipment risk distribution map through a visual interface.

[0192] Specific limitations regarding the intelligent inspection and early warning system for safety production integrating IoT data can be found in the above-mentioned limitations on the intelligent inspection and early warning methods for safety production integrating IoT data, and will not be repeated here. Each module in the aforementioned intelligent inspection and early warning system for safety production integrating IoT data can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0193] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0194] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for intelligent inspection and early warning of safe production that integrates Internet of Things (IoT) data, characterized in that, The intelligent inspection and early warning method for safe production that integrates IoT data includes: Acquire multi-source IoT monitoring data, including equipment operation vibration data, temperature data, and environmental status data; The multi-source IoT monitoring data is preprocessed at the edge to obtain preprocessed monitoring data; Based on the preprocessed monitoring data, multimodal state features are extracted, and a set of equipment operating state features is constructed. The set of equipment operating state features is then spatiotemporally aligned to obtain fused feature data under a unified reference coordinate. The fused feature data is input into a pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation result. Based on the equipment operation deviation results, a graded early warning determination is performed to obtain the early warning level, and the corresponding early warning handling strategy is triggered according to the early warning level.

2. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 1, characterized in that, The acquisition of multi-source IoT monitoring data specifically includes: Vibration data and temperature data of the equipment are collected according to a preset sampling frequency. The temperature data includes equipment operating temperature data and ambient temperature data. According to the preset inspection path and preset scanning angle, environmental status data of the preset inspection area is obtained. The environmental status data includes multispectral image data and environmental gas data. The device vibration data, temperature data, and environmental status data are uniformly packaged and processed to form the multi-source IoT monitoring data.

3. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 2, characterized in that, Prior to acquiring multi-source IoT monitoring data, the intelligent inspection and early warning method for safe production that integrates IoT data further includes: Acquire historical inspection data and historical operation records of the equipment, and determine the risk status information of the equipment based on the historical inspection data and the historical operation records of the equipment. The risk status information includes equipment risk level information and risk area distribution information. Based on the risk status information, priority inspection areas are determined, and the initial inspection path is adjusted based on the priority inspection areas to obtain the preset inspection path. The initial inspection path is a pre-set default inspection path. The initial sampling frequency of the corresponding monitoring node is adaptively adjusted based on the risk status information to obtain the preset sampling frequency, wherein the initial sampling frequency is a pre-set default sampling frequency.

4. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 1, characterized in that, The step of performing edge-side preprocessing on the multi-source IoT monitoring data to obtain preprocessed monitoring data specifically includes: The vibration data of the equipment operation is filtered to obtain preprocessed vibration data of the equipment operation. Thermal emissivity compensation is performed on the temperature data to obtain preprocessed temperature data; The environmental status data is subjected to image quality enhancement and gas data calibration processing to obtain environmental monitoring data; The preprocessed equipment operation vibration data, the preprocessed temperature data, and the environmental monitoring data are subjected to timestamp unification, data format conversion, and structured encapsulation to obtain the preprocessed monitoring data.

5. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 4, characterized in that, The filtering process for the equipment operating vibration data to obtain preprocessed equipment operating vibration data specifically includes: Obtain the device type, determine the characteristic frequency band of the vibration signal based on the device type, and dynamically determine the filter cutoff frequency based on the characteristic frequency band; The vibration data of the equipment operation is filtered according to the filter cutoff frequency to obtain the filtered vibration data. The filtered vibration data is then normalized to obtain normalized vibration data. The normalized vibration data is segmented according to a preset vibration feature extraction window to obtain the preprocessed equipment operation vibration data.

6. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 4, characterized in that, The process involves extracting multimodal state features from the preprocessed monitoring data, constructing a device operating state feature set, and performing spatiotemporal alignment processing on the device operating state feature set to obtain fused feature data under a unified reference coordinate system. Specifically, this includes: The frequency domain characteristic parameters of the vibration signal are calculated based on the preprocessed equipment operation vibration data to obtain the equipment operation status characteristics; Based on the preprocessed temperature data, the temperature distribution characteristics of the equipment operating area are constructed to obtain the thermal state characteristics of the equipment. Image region features and gas concentration distribution features are extracted from the environmental monitoring data to obtain the equipment environmental status features; The equipment operating status characteristics, the equipment thermal status characteristics, and the equipment environmental status characteristics are aggregated according to the equipment identifier to construct a set of equipment operating status characteristics. The device operating status feature set is spatiotemporally aligned to obtain fused feature data under a unified reference coordinate system.

7. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 1, characterized in that, The intelligent inspection and early warning method for safe production that integrates IoT data also includes: Acquire historical equipment operation monitoring data, and perform feature extraction processing on the historical equipment operation monitoring data to form a training sample set; The training sample set is labeled with status based on historical equipment operation records and historical fault records to obtain the training dataset. The anomaly detection model is iteratively trained using the training dataset to obtain the pre-built anomaly detection model, which includes a feature input layer, a state modeling layer, and a result output layer. During the training of the anomaly detection model, the model parameters are updated based on the loss function using a gradient optimization algorithm, which includes the Adam optimization algorithm.

8. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 1, characterized in that, The step of performing a graded early warning determination based on the equipment operation deviation result to obtain an early warning level, and triggering a corresponding early warning handling strategy based on the early warning level, specifically includes: The equipment operating status level is determined by comparing the equipment operating deviation result with a preset health threshold range. Map the device's operating status level to the corresponding warning level; According to the warning level, a corresponding warning handling strategy is matched from the preset warning strategy library and executed. The warning handling strategy includes at least one of on-site prompt warning, remote notification warning and safety linkage control.

9. The intelligent inspection and early warning method for safe production integrating IoT data according to claim 1, characterized in that, The intelligent inspection and early warning method for safe production that integrates IoT data also includes: Based on the equipment operation deviation results and early warning levels, generate equipment risk distribution information; The equipment risk distribution information is mapped to the spatial layout of the inspection area to construct an equipment risk distribution map, which is then displayed through a visual interface.

10. A smart inspection and early warning system for safe production that integrates Internet of Things (IoT) data, characterized in that: The intelligent safety inspection and early warning system that integrates IoT data includes: The monitoring data acquisition module is used to acquire multi-source IoT monitoring data, which includes equipment operation vibration data, temperature data, and environmental status data. An edge preprocessing module is used to perform edge-side preprocessing on the multi-source IoT monitoring data to obtain preprocessed monitoring data; The multimodal feature construction module is used to extract multimodal state features based on the preprocessed monitoring data, construct a set of equipment operating state features, and perform spatiotemporal alignment processing on the set of equipment operating state features to obtain fused feature data under a unified reference coordinate. An anomaly assessment module is used to input the fused feature data into a pre-built anomaly identification model to assess the health status and obtain the equipment operation deviation result. The early warning determination module is used to perform graded early warning determination based on the equipment operation deviation results, obtain the early warning level, and trigger the corresponding early warning handling strategy according to the early warning level.

Images