Method, device and equipment for checking hidden danger of gas drainer and medium

Abstract

The invention discloses a gas drainer hidden danger checking method, device and equipment and a medium, and belongs to the technical field of metallurgy. The method comprises the following steps: acquiring an image and operation state data of the gas drainer, wherein the operation state data at least comprises one of temperature, pressure, liquid level and carbon monoxide concentration; determining a first health score of the gas drainer according to the image; determining a second health score of the gas drainer according to the operation state data; determining a comprehensive health score of the gas drainer according to the first health score and the second health score; and determining the hidden danger grade of the gas drainer according to the comprehensive health score. According to the method, the comprehensive health score of the gas drainer is objectively and accurately obtained by automatically and comprehensively analyzing the image and the operation state data of the gas drainer without manual participation, so that whether the gas drainer has a hidden danger or not can be checked according to the comprehensive health score.

Description

Technical Field

[0001] The present invention relates to the technical field of metallurgy, and in particular, to a method, device, equipment and medium for troubleshooting potential hazards of a gas drainer. Background Art

[0002] The gas drainer is a key device in the gas transmission and distribution system of iron and steel enterprises. Its main function is to timely drain the condensed water in the gas pipeline to prevent the accumulated water in the pipeline from having an adverse impact on the gas transmission efficiency and equipment safety.

[0003] In the prior art, the operation status of the gas drainer mainly depends on manual inspection. The inspection personnel record the operation situation of the equipment through on-site inspection and conduct potential hazard investigation according to relevant operation specifications. This manual inspection method combines the experience of the inspection personnel and the operation specifications, and can ensure the safe operation of the gas drainer to a certain extent.

[0004] However, with the expansion of the scale of iron and steel enterprises and the increase in the number of gas drainers, the traditional manual inspection mode not only has a large workload, low efficiency and long time consumption, but also the inspection of key points is restricted by time and human resources, resulting in some potential hazards being missed. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a method, device, equipment and medium for troubleshooting potential hazards of a gas drainer, which can automatically obtain the comprehensive health score of the gas drainer objectively and accurately through comprehensive analysis of the image and operation status data of the gas drainer, so as to find out whether there are potential hazards in the gas drainer according to the comprehensive health score.

[0006] In a first aspect, the present invention provides a method for troubleshooting potential hazards of a gas drainer, the method comprising:

[0007] Obtaining the image and operation status data of the gas drainer, where the operation status data includes at least one of temperature, pressure, liquid level and carbon monoxide concentration;

[0008] Determining a first health score of the gas drainer according to the image;

[0009] Determining a second health score of the gas drainer according to the operation status data;

[0010] Determining a comprehensive health score of the gas drainer according to the first health score and the second health score;

[0011] Determining the potential hazard level of the gas drainer according to the comprehensive health score.

[0012] Optionally, determining the first health score of the gas drainer according to the image includes:

[0013] Inputting the image into a pre-built neural network model to obtain the first health score output by the neural network model;

[0014] Wherein, the neural network model is trained by an image set of the gas drainer, and the image set includes normal operation images and abnormal operation images.

[0015] Optionally, determining the second health score of the gas drainer according to the operation status data includes:

[0016] Inputting the operation status data into a pre-built memory network model to obtain the second health score output by the memory network model;

[0017] Wherein, the memory network model is trained by an operation status data set of the gas drainer, and the operation status data set includes normal operation status data and abnormal operation status data.

[0018] Optionally, determining the comprehensive health score of the gas drainer according to the first health score and the second health score includes:

[0019] Obtaining a first weight of the first health score and a second weight of the second health score;

[0020] Calculating a first product of the first health score and the first weight, and a second product of the second health score and the second weight;

[0021] Calculating the sum of the first product and the second product to obtain the comprehensive health score of the gas drainer.

[0022] Optionally, determining the hidden danger level of the gas drainer according to the comprehensive health score includes:

[0023] If the comprehensive health score is less than or equal to a preset first threshold, determining that the hidden danger level of the gas drainer is a high level;

[0024] If the comprehensive health score is greater than the first threshold and less than a preset second threshold, the hidden danger level of the gas drainer is a medium level;

[0025] If the comprehensive health score is greater than or equal to the second threshold, the hidden danger level of the gas drainer is a low level.

[0026] Optionally, the method further includes:

[0027] Obtaining the false alarm rate of the hidden danger of the gas drainer;

[0028] Adjust at least one of the first threshold and the second threshold according to the false alarm rate of the potential hazard.

[0029] Optionally, a plurality of the gas drainers are included, and the method further includes:

[0030] Determine the confidence level of the operation status data of each gas drainer;

[0031] Add the operation status data with the confidence level greater than a preset confidence level threshold to the operation status data set to obtain a new operation status data set;

[0032] Optimize the memory network model using the new operation status data set.

[0033] In a second aspect, the present invention provides a device for detecting potential hazards of a gas drainer, and the device includes:

[0034] An acquisition module, configured to acquire an image and operation status data of the gas drainer, where the operation status data includes at least one of temperature, pressure, liquid level, and carbon monoxide concentration;

[0035] A first determination module, configured to determine a first health score of the gas drainer according to the image;

[0036] A second determination module, configured to determine a second health score of the gas drainer according to the operation status data;

[0037] A third determination module, configured to determine a comprehensive health score of the gas drainer according to the first health score and the second health score;

[0038] A fourth determination module, configured to determine the potential hazard level of the gas drainer according to the comprehensive health score.

[0039] In a third aspect, the present invention provides an electronic device, including: a memory and a processor, where the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the method as described in the first aspect.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the method as described in the first aspect.

[0041] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:

[0042] A method, device, equipment and medium for troubleshooting potential hazards of a gas drainer provided by an embodiment of the present invention obtain an image and operating status data of the gas drainer to understand the appearance status and internal operating status of the gas drainer. The operating status data includes at least one of temperature, pressure, liquid level and carbon monoxide concentration; according to the image, determine the first health score of the gas drainer to evaluate the appearance health condition; according to the operating status data, determine the second health score of the gas drainer to evaluate the internal operating status; according to the first health score and the second health score, determine the comprehensive health score of the gas drainer to understand the comprehensive health condition of the gas drainer; according to the comprehensive health score, determine the potential hazard level of the gas drainer to identify potential hazards of the gas drainer. This method automatically obtains the comprehensive health score of the gas drainer objectively and accurately through comprehensive analysis of the image and operating status data of the gas drainer without manual participation, so that it can be determined whether there are potential hazards in the gas drainer according to the comprehensive health score.

[0043] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically illustrates the specific embodiments of the present invention. Brief Description of the Drawings

[0044] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. And throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0045] Figure 1 is a flowchart of a method for troubleshooting potential hazards of a gas drainer provided by an embodiment of the present invention;

[0046] Figure 2 is a structural schematic diagram of a gas drainer provided by an embodiment of the present invention;

[0047] Figure 3 is a structural block diagram of a device for troubleshooting potential hazards of a gas drainer provided by an embodiment of the present invention. Detailed Embodiments

[0048] To make the objectives, technical solutions and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the drawings. It should be understood that the embodiments of the present disclosure and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0049] Figure 1 is a flowchart of a method for detecting potential hazards of a gas drainer provided by an embodiment of the present invention. As Figure 1 shown, the method includes:

[0050] Step S110: Obtain the image and operation status data of the gas drainer.

[0051] Among them, the operation status data includes at least one of temperature, pressure, liquid level, and carbon monoxide concentration.

[0052] Figure 2 is a schematic structural diagram of a gas drainer provided by an embodiment of the present invention. As Figure 2 shown, the gas drainer is connected to the gas pipeline. The gas drain valve includes a primary valve 1, a connecting pipe 2, a detection port 3, a secondary valve 4, a flange port 5, a water filling port 6, an exhaust port 7, a plug 8, a cylinder body 9, an overflow port 10, a drain port 11, a sewage handhole 12, a sewage valve 13, and a drain valve 14.

[0053] In the embodiment of the present application, video inspection can be performed through a video inspection platform according to a preset operation path to automatically capture images of key points of the gas drainer, and then obtain the images captured by the video inspection platform through the Real Time Streaming Protocol (RTSP). Among them, the key points may include images of parts such as the primary valve 1, the connecting pipe 2, the detection port 3, the secondary valve 4, the flange port 5, the water filling port 6, the exhaust port 7, the plug 8, the cylinder body 9, the overflow port 10, the drain port 11, the sewage handhole 12, the sewage valve 13, and the drain valve 14. The images may include not only the images of the key points, but also the images of the power supply line, the power supply box, and the surrounding environment of the gas drainer.

[0054] Among them, the images captured by the video inspection platform can be denoised, cropped, and format-adjusted to obtain images of the gas drainer, improving the visual quality of the images. For example, a Gaussian filter is used to denoise the images captured by the video inspection platform to reduce the noise caused by environmental interference and obtain smooth images. The denoising principle is to replace the pixel value with the weighted average of its surrounding pixel values, effectively reducing the impact of random noise while retaining the main structural information of the image. The kernel size of the filter can be adjusted according to the image resolution and noise intensity. Exemplarily, a 3×3 or 5×5 filter kernel can be selected. During the image cropping process, combined with the equipment characteristics of the gas drainer and the specific requirements of the monitoring points, the regions of interest in the image are extracted. These regions include key points such as valves, drain ports, and flange connections, which can reduce the interference of irrelevant regions on the analysis results. The boundaries of the regions of interest (ROI) can also be determined by an automatic positioning algorithm or a manual calibration method. For example, edge detection algorithms are used in combination with target feature templates for automatic region extraction.

[0055] In addition, the brightness and contrast of the cropped images are adjusted to ensure that image features can be clearly captured in low-light or high-contrast environments. It can be understood that these image processing steps can effectively improve the image quality and provide better input data for subsequent steps.

[0056] Exemplarily, the video inspection platform adopts an Artificial Intelligence (AI) video image analysis system to achieve online automatic video inspection.

[0057] Among them, the video inspection platform includes a camera, which is installed above the peripheral guardrail of the gas drainer. Edge computing technology can be combined to preprocess the video images captured by the camera. For example, an edge computing module is embedded in the video monitoring device to detect simple abnormal features (such as obvious leakage or structural damage) in the video stream in real time, and the processed results are transmitted to the backend server together with the original images. This can improve the data utilization efficiency while reducing the amount of transmitted data. In some special environments, such as high-temperature, high-humidity, or high-gas-concentration areas, cameras with explosion-proof, anti-fog, and autofocus functions can be selected to ensure the safety of the inspection and the image quality. Exemplarily, an industrial-grade camera supporting 1080P resolution can be used, with a focal length range between 10-100 millimeters, which can meet the needs of long-distance and short-distance monitoring.

[0058] It is understandable that during the automatic inspection process, that is, during the process of automatically capturing images, it is necessary to perform real-time scheduling on the operations of the camera to ensure the accuracy and efficiency of the inspection. The pan-tilt of the camera is remotely controlled through the ONVIF (Open Network Video Interface Forum) protocol, specifically including horizontal rotation, vertical adjustment, and lens zoom operations. The adjustment parameters of the camera can be set according to the specific location of the monitoring point and the characteristics of the equipment. For example, for the overflow port that is far from the camera, the focal length of the lens can be increased to obtain a clearer image; while for the components such as valves and flange interfaces at close range, the field of view can be enhanced by reducing the focal length.

[0059] To further improve the intelligent level of troubleshooting, it also supports dynamically adjusting the inspection sequence and frequency of the camera. For example, when potential hazards frequently occur at a certain monitoring point, inspections of this point will be prioritized and its inspection frequency will be increased; conversely, for monitoring points that have not shown abnormalities for a long time, their inspection frequency can be appropriately reduced. This dynamic adjustment mechanism can optimize the allocation of inspection resources and improve the inspection efficiency.

[0060] In the embodiment of the present application, temperature data can be collected through a temperature sensor, pressure data through a pressure sensor, liquid level data through a liquid level sensor, and carbon monoxide concentration data through a carbon monoxide concentration sensor. The data collected by each sensor is obtained through the Message Queuing Telemetry Transport (MQTT) protocol. This protocol has low power consumption, lightweight, and efficient data transmission capabilities, making it suitable for multi-sensor data collection in an industrial environment. Among them, the sensor sampling frequency can be set to once per second, which can not only reflect the state changes of the gas drainer in real time but also effectively reduce data redundancy. The reliability and flexibility of the MQTT protocol ensure the synchronous collection and stable transmission of data from multiple sensors.

[0061] In the embodiment of the present application, the data collected by each sensor can be obtained at a set frequency to obtain the time-series data collected by each sensor. Then, after processing such as normalizing and cleaning the time-series data collected by each sensor, the operating state data of the gas drainer is obtained, effectively improving the quality, consistency, availability, integrity, and analysis efficiency of the data, providing a data basis for the reliable execution of subsequent steps. The normalization formula is as follows:

[0062]

[0063] where x ′ is the operating state data, x max and x minThey are the minimum and maximum values in the time series data respectively. This linear normalization method can map data in different ranges to the interval [0, 1], facilitating subsequent analysis and calculation. For some sensor data (such as liquid level data), a logarithmic normalization method can also be used to compress the dynamic range of the data.

[0064] In the embodiments of the present application, the time series data can also be cleaned. For example, the data at consecutive time points can be smoothed by a mean filtering method to suppress the influence of sudden noises.

[0065] Exemplarily, the following formula is used to update the data at the current time point:

[0066]

[0067] where y t is the smoothed data, x is the data at the current time point and its previous n - 1 time points, and n is the window size. The window size can be selected according to the fluctuation characteristics of the data, and common values are 3 or 5.

[0068] To eliminate outliers, the embodiments of the present application adopt a statistical analysis method. For example, the mean and standard deviation of the sensor data are calculated, and the abnormal data that significantly deviates from the normal value is eliminated by setting the upper and lower threshold ranges. Specifically, a threshold range can be defined as:

[0069] [μ - kσ, μ + kσ];

[0070] where μ is the mean of the data, σ is the standard deviation of the data, and k is an adjustment factor. The threshold range is from 2 to 3, and the values exceeding this range are regarded as abnormal data and eliminated.

[0071] It should be noted that in some embodiments, to further improve the processing accuracy of the time series data, an anomaly detection algorithm based on machine learning, such as Isolation - Forest or Support Vector Machine (SVM), is adopted to identify the abnormal patterns of the sensor data. These algorithms can more accurately calibrate and eliminate outliers in complex environments.

[0072] In the embodiments of the present application, to improve the acquisition accuracy and anti - interference ability of the time series data, the sensor signal is digitally acquired after hardware filtering. For example, a low - pass filter can be introduced into the circuit of the liquid level sensor to suppress the influence of environmental noise on the measurement result. For the pressure sensor, an industrial - grade device with an accuracy of 0.1% FS (full scale) can be selected to ensure the reliability of the measurement.

[0073] In the embodiments of the present application, temperature characterizes the thermodynamic properties of the internal operating state of the gas drainer; pressure characterizes the safety of the gas drainer pipeline and the internal system; liquid level characterizes the change of the water level inside the gas drainer to prevent abnormal water accumulation or insufficient water level; carbon monoxide concentration characterizes the potential risk of gas leakage and is an important indicator for the safety monitoring of the gas drainer. Therefore, the operating state data can reflect the internal operating state of the gas drainer, such as pressure fluctuations and liquid level conditions, etc.

[0074] Among them, each sensor is evenly arranged at the key component positions of the gas drainer. For example, the pressure sensor and the liquid level sensor can be installed at the inlet and the bottom of the gas drainer respectively to ensure that the collected data can comprehensively reflect the operating state of the equipment.

[0075] It can be understood that data processing, as a key step, significantly improves the quality and consistency of the data through denoising, cropping, and brightness adjustment of image data, as well as normalization, smoothing, and outlier cleaning of time-series data, providing high-quality input for subsequent steps. At the same time, the application of these processing technologies can adapt to the characteristics of different types of gas drainer equipment and the differences in the working environment, thus enhancing the versatility and stability of the system.

[0076] Step S120: Determine the first health score of the gas drainer according to the image.

[0077] In the embodiments of the present application, the image can reflect features such as the texture, shape, and damage marks of the appearance of the gas drainer. Therefore, the external health condition of the gas drainer can be reflected by the first health score determined according to the image.

[0078] Step S130: Determine the second health score of the gas drainer according to the operating state data.

[0079] In the embodiments of the present application, the operating state data can reflect the dynamic changes of the state of the gas drainer. Therefore, whether the internal operating state of the gas drainer is normal can be reflected by the second health score determined according to the operating state data.

[0080] Step S140: Determine the comprehensive health score of the gas drainer according to the first health score and the second health score.

[0081] In the embodiments of the present application, through multi-modal fusion algorithm analysis based on the appearance condition and the actual operating condition of the gas drainer, the comprehensive health score that comprehensively reflects the gas drainer is obtained, that is, the health condition of the gas drainer is reflected by the comprehensive health score.

[0082] Step S150: Determine the hidden danger level of the gas drainer according to the comprehensive health score.

[0083] In the embodiments of the present application, the hidden danger level of the gas drainer is determined according to the health condition of the gas drainer. For example, if the gas drainer is relatively healthy, the hidden danger level is relatively low; if the gas drainer is unhealthy, the hidden danger level is relatively high.

[0084] In the embodiments of the present application, a remote automatic inspection system can be constructed by combining AI technology with multi-modal data analysis. The method in the embodiments of the present application can be applied to this remote automatic inspection system to achieve comprehensive monitoring of key points inside and outside the gas drainer, be able to grasp the operating state and appearance of the gas drainer in real time, and predict the level of potential hidden dangers. The system can automatically generate curves and analysis reports of the operating state data, give early warnings for hidden dangers, provide decision-making basis for operation and maintenance personnel. The all-weather monitoring and data analysis not only improve the accuracy of hidden danger identification, but also further optimize the equipment maintenance plan, reduce the failure rate and maintenance cost. This method can fully automatically check the hidden danger level of the gas drainer, minimize manual intervention, significantly reduce the necessity for operation and maintenance personnel to enter the gas dangerous area, and fundamentally reduce the occurrence probability of potential accidents such as poisoning, asphyxiation, slipping and falling, vehicle injury, and falling from height of operation and maintenance personnel.

[0085] In the embodiments of the present application, a corresponding real-time alarm function is set according to the hidden danger level, and the real-time alarm is prompted by means of sound, light, and platform pop-up window. Sound alarm: A buzzer sound is emitted through the on-site alarm device to alert the staff. Light alarm: Alarm lights are set in the monitoring area, and when a hidden danger is identified, the alarm lights will flash. Platform pop-up reminder: A hidden danger alarm window pops up on the management platform of the monitoring center, and at the same time, specific hidden danger information is displayed.

[0086] It should be noted that the early warning content usually includes the following information:

[0087] Hidden danger location: Clearly point out the specific components and monitoring points of the gas drainer.

[0088] Hidden danger type: Such as flange leakage, sewage outlet blockage, etc.

[0089] Suggested treatment measures: Provide treatment suggestions according to the hidden danger type, such as replacing the gasket, cleaning the sewage outlet, etc.

[0090] In the embodiments of the present application, there are multiple gas drainers, which are distributed along the gas pipeline. Each gas drainer has a corresponding number, and the hidden danger level of each gas drainer can be recorded one-to-one with the corresponding number in the hidden danger report. The following contents are also recorded in the hidden danger report: the location of the hidden danger, the type of the hidden danger, and the recommended treatment measures. This method has the functions of all-day real-time monitoring and abnormal alarm, can quickly locate and prompt for treatment when a hidden danger occurs, effectively avoid accidents, significantly improve the intrinsic safety of the gas drainer area, and help achieve the management goal of unattended operation.

[0091] In one implementation, the hidden danger report also includes all health scores, hidden danger detection results, and alarm records of this inspection. Exemplarily, the records are uploaded to the background database for managers to query and analyze at any time.

[0092] The embodiments of the present application complete the active monitoring and intelligent identification of the hidden dangers of gas drainers by combining the remote control of cameras, the real-time application of multi-modal analysis, and the comprehensive implementation of the early warning mechanism. It can not only accurately capture potential problems in the operation of gas drainers, but also promptly send an alarm to managers, providing an important basis for subsequent hidden danger handling, thus effectively ensuring the safe operation of gas drainers.

[0093] Among them, this method can detect each key point of the gas drainer, effectively check key areas, avoid missed inspections and omissions. Compared with the traditional manual inspection method, this method greatly shortens the inspection time, and at the same time ensures the accuracy and coverage of the inspection results, greatly improving the inspection efficiency.

[0094] Optionally, step S120 includes:

[0095] Input the image into a pre-built neural network model to obtain the first health score output by the neural network model.

[0096] Among them, the neural network model is trained by an image set of gas drainers, and the image set includes normal operation images and abnormal operation images.

[0097] In the embodiments of the present application, a large number of normal operation images and abnormal operation images of gas drainers can be collected to form an image set, and the initial neural network model is trained using the image set to obtain the neural network model. Among them, the initial neural network model can be a convolutional neural network model (Convolutional Neural Networks, CNN).

[0098] Specifically, a neural network model can be used to analyze images, extract key features of the appearance of the gas drainer, and determine whether there are problems with the power supply line, power supply box, thermal insulation, tracing tape, cylinder body, and valve according to the key features. For example, whether the valve is corroded, whether the interface leaks, and whether the valve has physical damage. Thus, the structural health of the gas drainer can be scored according to the problems that occur, and a first health score can be obtained. Through the video platform and the neural network model, online automatic video inspection, accurate hidden danger investigation, and early warning can be realized, improving the efficiency of the investigation work. This not only eliminates the safety risks brought by inspectors going to the site but also reduces the risks of poisoning and asphyxiation accidents caused by insufficient inspection of gas drainers by humans based on experience, further improving the essential management level of gas drainers.

[0099] Through the operations of the convolutional layer and pooling layer, the CNN extracts and expresses the texture, shape, and edge features of the input image at multiple levels, and finally completes the hidden danger classification in the fully connected layer.

[0100] As an implementation method, the input of the CNN is the preprocessed image, and the output is the multi-classification result or the health score. The multi-classification result can include hidden danger types such as "normal", "corroded", and "leaking". The core calculation formula of the convolutional layer is:

[0101] y i,j,k =f(∑ m,n x i+m,j+n ·w m,n,k +b k );

[0102] Among them, y i,j,k represents the i, j position of the Kth channel in the output feature map of the convolution result, x i+m,j+n is the pixel value of the input image at the corresponding position, w m,n,k is the convolution kernel weight, b k is the bias, and f is the activation function (usually the ReLU function is selected). It should be noted that through multiple convolutional operations, the CNN can extract high-order features in the gas drainer image layer by layer, so as to achieve accurate identification of hidden dangers.

[0103] In some embodiments, to further enhance the robustness of the analysis, an anomaly detection module is also introduced to check the consistency of the output results of the LSTM and the CNN. For example, when there are significant differences between the LSTM analysis result and the CNN analysis result, the corresponding key points will be marked as "to be rechecked".

[0104] Optionally, step S130 includes:

[0105] Input the operating state data into the pre-built memory network model to obtain the second health score output by the memory network model.

[0106] Among them, the memory network model is trained by the operation status data set of the gas drainer, and the operation status data set includes normal operation status data and abnormal operation status data.

[0107] In the embodiment of the present application, a large amount of normal operation status data and abnormal operation status data of gas drainers can be collected to form an operation status data set, and then the initial memory network model is trained by the operation status data set to obtain the memory network model. Among them, the initial memory network model can be a Long Short-Term Memory (LSTM) model. LSTM is a neural network structure that can capture long-term dependencies and is particularly suitable for processing time-correlated sequence data such as sensor data. Specifically, the LSTM model screens and stores the input data through a gating mechanism (including an input gate, a forget gate, and an output gate) to learn the trend and periodic characteristics of the data changing over time. For example, for the pressure data of the gas drainer, the LSTM model can identify the pattern of pressure changes and capture the potential hazard signals corresponding to abnormal fluctuations.

[0108] Specifically, the operation status data can be analyzed through the memory network model to obtain the dynamic trend of the operation status of the gas drainer, identify whether there are potential fault hazards, and then score the internal health of the gas drainer according to the identified situation to obtain the second health score.

[0109] As an option, the input of the LSTM model is a multi-dimensional time series data sequence after normalization, such as temperature, pressure, liquid level, and carbon monoxide concentration, etc., and the output is the second health score (Health-Score) corresponding to each time point. The specific calculation process of the model includes the following formula:

[0110] h t =f(W x ·x t +W h ·h t-1 +b);

[0111] Among them, h t represents the hidden state at the current time point, x t is the input data at the current time point, h t-1 is the hidden state at the previous time point, W x and W h are the input weight matrix and the hidden state weight matrix respectively, b is the bias, and f is the activation function (usually the Tanh function is selected). It should be noted that the LSTM model can dynamically learn the characteristics of time series data through the above formula, and then identify the operation status of the gas drainer.

[0112] Optionally, there are multiple gas drainers, and the method further includes:

[0113] Determine the confidence level of the operation status data of each gas drainer; add the operation status data with a confidence level greater than the preset confidence level threshold to the operation status data set to obtain a new operation status data set; use the new operation status data set to optimize the memory network model.

[0114] In the embodiments of the present application, the memory network model will also be optimized through the active learning and optimization module to further improve the intelligence level. Among them, experts review and annotate the newly collected operation status data, screen out the operation status data with a confidence level greater than the confidence level threshold as new samples, and then perform incremental training on the memory network model in combination with the original operation status data set, dynamically adjusting the model parameters to adapt to environmental changes and the emergence of new hidden dangers. At the same time, the confidence level threshold can be adjusted according to the actual detection situation to ensure that hidden dangers are not missed while reducing false alarms. Through the active learning mechanism, the model can continuously optimize itself to achieve the accuracy and robustness of hidden danger detection.

[0115] In the embodiments of the present application, the same idea can be used to optimize the neural network model to improve the accuracy and robustness of hidden danger detection.

[0116] Among them, the following loss function is used in the model optimization process:

[0117]

[0118] Among them, y i is the actual value, f(x i ) is the predicted value, R(f) is the regularization term, and λ is the regularization coefficient.

[0119] Active learning and model optimization are important links to realize the self-iteration and optimization of the model. Based on hidden danger identification and handling, through expert feedback and hidden danger handling records, the deep learning models (including LSTM and CNN models) for multi-modal data analysis are continuously optimized, so as to improve the adaptability to new hidden dangers and the detection accuracy. Active learning continuously improves the generalization ability of the model by introducing high-confidence operation status data and newly annotated data into the training set, ensuring the stability of the method for investigation under different environments and working conditions.

[0120] In one implementation, the active learning mechanism completes model optimization through the following steps:

[0121] Data Screening: Extract representative high-confidence samples from the hidden danger handling records, such as samples with obvious anomalies and successful verification of handling results; Dynamically adjust the confidence threshold for hidden danger detection according to the historical false alarm rate and missed alarm rate of the model; For example, for some low-risk hidden dangers, the confidence threshold can be appropriately reduced to improve detection sensitivity; For high-risk hidden dangers, the confidence threshold needs to be increased to avoid false alarm interference. It should be noted that the dynamic adjustment of confidence can significantly improve the stability and reliability of the model, especially under complex working conditions.

[0122] Feature Enhancement: Perform data enhancement processing on the labeled data, such as rotating, scaling, adding noise, etc. to image data, to improve the robustness of the model to different scenarios.

[0123] Model Update: Optimize the LSTM and CNN models respectively based on the incremental training method to ensure the synchronous improvement of the time series data analysis and image data analysis capabilities.

[0124] To further improve the efficiency of active learning, the embodiment of the present application also introduces the feature correlation analysis of multi-modal data. For example, by combining the time series features output by the LSTM model and the image features output by the CNN model, pattern matching is performed on abnormal hidden dangers, and the key feature vectors of the hidden dangers are generated to guide the direction of model optimization. Exemplarily, the feature vectors may include the following:

[0125] The trend change features in the time series data, such as the pressure fluctuation pattern;

[0126] The local damage features in the image data, such as the crack length or corrosion area.

[0127] In some embodiments, it supports clustering analysis of the hidden danger types marked by experts, so as to discover potential patterns in the hidden danger data. For example, through clustering analysis, the occurrence frequency and typical features of certain specific hidden dangers can be identified, and these hidden danger types can be intensively trained during model optimization. It also supports the visual display of the model optimization process, such as showing the dynamic changes of the training loss, verification accuracy, and confidence threshold of the model through a curve graph, which is convenient for managers to monitor the optimization progress in real time.

[0128] It should be noted that "active learning and model optimization" realizes the continuous optimization of the hidden danger detection model by introducing high-confidence samples into model training and combining technical means such as incremental training, confidence adjustment, and multi-modal feature analysis. This optimization process not only enhances the adaptability of the model to complex working conditions, but also significantly improves the accuracy and efficiency of hidden danger identification, providing strong technical support for the intelligent inspection of gas drainers.

[0129] Optionally, step S140 includes:

[0130] Obtain the first weight of the first health score and the second weight of the second health score; calculate the first product of the first health score and the first weight, and the second product of the second health score and the second weight; calculate the sum of the first product and the second product to obtain the comprehensive health score of the gas drainer.

[0131] In the embodiment of the present application, the analysis results of the LSTM model and the CNN model are integrated through a multimodal fusion algorithm to generate an overall operating state evaluation of the gas drainer, that is, a comprehensive health score. The multimodal fusion adopts a weighted fusion strategy, and its fusion formula is:

[0132] F fusion =αF LSTM +βF CNN ;

[0133] Where, F fusion represents the comprehensive health score, F LSTM represents the second health score, F CNN represents the first health score, α represents the first weight, β represents the second weight, and α + β = 1.

[0134] Dynamically adjust the first weight and the second weight according to the accuracy of the operating state data. For example, when the stability of the sensor data is relatively high, the second weight can be appropriately increased. The introduction of the fusion algorithm can simultaneously consider the internal state (such as time series characteristics of temperature, pressure, etc.) and external features (such as image texture and structural damage) of the gas drainer, thereby significantly improving the comprehensiveness and accuracy of hidden danger detection.

[0135] Optionally, step S150 includes:

[0136] If the comprehensive health score is less than or equal to a preset first threshold, determine that the hidden danger level of the gas drainer is a high level; if the comprehensive health score is greater than the first threshold and less than a preset second threshold, the hidden danger level of the gas drainer is a medium level; if the comprehensive health score is greater than or equal to the second threshold, the hidden danger level of the gas drainer is a low level.

[0137] In the embodiment of the present application, the hidden danger levels of the gas drainer are classified, and the hidden danger levels can include low level, medium level, and high level. Among them, high-level hidden dangers need to be dealt with in a timely manner, otherwise safety accidents may occur. The handling time limit for medium-level hidden dangers can be longer, and low-level hidden dangers can be observed or not dealt with.

[0138] Optionally, the method further includes:

[0139] Obtain the false alarm rate of the hidden dangers of the gas drainer; adjust at least one of the first threshold and the second threshold according to the false alarm rate of the hidden dangers.

[0140] In the embodiments of the present application, a warning will also be issued after a potential hazard is discovered, and it specifically relies on the following adjustment rules:

[0141] (1) Dynamically adjust the first threshold and / or the second threshold according to the type of potential hazard point and the historical false alarm rate;

[0142] (2) Automatically mark high-level and low-level potential hazard points as "pending review";

[0143] (3) As the model is optimized, gradually increase the confidence threshold to reduce false alarms.

[0144] In the embodiments of the present application, the method further includes:

[0145] A. Automatically generate potential hazard handling tasks and assign the potential hazard handling tasks to the corresponding responsible persons;

[0146] B. Set a time limit for handling potential hazards. If the handling is not completed within the time limit, a second reminder will be issued;

[0147] C. After the potential hazard handling is completed, the responsible person submits the handling result for acceptance and archiving, forming a closed-loop management.

[0148] Among them, an inspection report can be generated according to the potential hazard report, and the inspection report includes the following information:

[0149] (1) The number and location identification of the gas drainer, used to clarify the equipment to which the potential hazard belongs;

[0150] (2) The detection time, specific location and detailed description of the potential hazard point, such as "flange interface leakage" or "valve not fully locked";

[0151] (3) The responsible person for handling the potential hazard and the handling period, so as to facilitate the responsible person to respond quickly;

[0152] (4) The re-inspection record after the potential hazard handling is completed;

[0153] (5) The record of multiple alarm times and the number of reminders when the potential hazard handling is delayed.

[0154] Specifically, closed-loop management is a key step to realize the whole-process tracking and management of potential hazard handling. The main purpose is that after a potential hazard is automatically identified, a potential hazard report is generated by the system, the handling task is automatically assigned, the handling progress is monitored in real time, and the handling result is accepted and archived, forming a closed-loop management process. It should be noted that the comprehensiveness and systematicness of closed-loop management not only ensure the high efficiency of potential hazard treatment, but also provide a reliable basis for subsequent potential hazard analysis and method optimization, and use automation and informatization technologies to realize the digitization, transparency and traceability of potential hazard handling.

[0155] Among them, the potential hazard handling task assignment rules can include the following dimensions:

[0156] Assign tasks to the person in charge of the area according to the area or equipment to which the potential hazard point belongs; for high-risk potential hazards, give priority to assigning them to senior technical personnel or management personnel; if a certain person in charge has a large number of current tasks, automatically adjust the task assignment order to optimize resource utilization.

[0157] Exemplarily, when assigning tasks, the task information will be sent to the person in charge by means of text messages, emails or platform notifications. The task information includes the content of the potential hazard report, the handling time limit and the handling requirements.

[0158] Specifically, set a time limit mechanism for potential hazard handling to ensure that potential hazards can be handled in a timely manner. For example, the handling time limit for high-risk potential hazards can be set to 24 hours, medium-risk potential hazards to 72 hours, and low-risk potential hazards to one week. Send a reminder notice before the handling time limit expires. If the task is not completed on time, a second reminder will be issued, and the overtime situation will be recorded.

[0159] It should be noted that after the potential hazard handling is completed, the person in charge needs to submit the handling result through the system and upload relevant handling records or supporting documents. For example, if the potential hazard is the leakage of a flange interface, the handling record can include photos of the replacement of the gasket and operation instructions. Automatically verify the submitted handling result, such as comparing the data changes before and after handling, or manually review by the management personnel. After the handling result is verified and passed, file the handling record to form a complete closed-loop management record of potential hazards.

[0160] In some embodiments, to further improve the standardization and transparency of management, support real-time monitoring of the potential hazard handling process and query of historical records. For example, the management personnel can view the handling progress of all potential hazard tasks through the monitoring platform, including the number and specific information of completed, in-progress and overtime tasks. In addition, statistical reports on potential hazard handling can be generated for analyzing the distribution of potential hazards, handling efficiency and personnel performance.

[0161] In order to achieve the closed-loop of the potential hazard handling process, a mechanism of multiple reminders and re-inspections is also introduced. For example, for some potential hazards that need to be monitored for a long time, regularly remind the person in charge to conduct re-inspections and record the re-inspection results. If it is found during the re-inspection that the potential hazard handling is not thorough, a new potential hazard task will be generated and a new round of closed-loop process will be started.

[0162] It can be understood that "closed-loop management" is not only the end link of potential hazard handling, but also an important source of data accumulation and knowledge update. The archived potential hazard handling records provide key data support for subsequent trend analysis and model optimization. For example, through the statistical analysis of historical potential hazard records, the occurrence frequency of certain potential hazards can be predicted, and the inspection path and frequency of cameras can be dynamically adjusted. In addition, the archived data can also be used for training new employees to improve their inspection and potential hazard handling capabilities.

[0163] The embodiments of the present application also support cross - departmental collaborative management functions. For example, for complex potential hazards, the system can simultaneously assign tasks to multiple responsible persons and provide a task collaboration platform for different departments to handle them collaboratively. After the tasks are completed, the collaboration platform will automatically record the contributions of each participant for subsequent assessment.

[0164] It should be noted that "closed - loop management" constructs a complete closed - loop for potential hazard treatment through the tracking and management of the whole process of potential hazard handling, which not only improves the efficiency and transparency of potential hazard handling, but also provides data support for long - term optimization. By closely integrating the links of task assignment, handling, acceptance, and filing, the standardization and process of potential hazard handling are realized, effectively ensuring the safe operation of the gas drainer.

[0165] The active learning mechanism relies on the potential hazard handling records in the closed - loop management and the manual annotation results of experts. Specifically, when a potential hazard is detected, the expert will review the potential hazard report generated by the system and annotate the accuracy of the detection result. The annotation results are stored as high - confidence samples and used to update the training set of the model. These samples can include various types of potential hazards, such as valve leakage, flange corrosion, sewage outlet blockage, etc., to cover all potential hazard scenarios that may occur during the operation of the gas drainer.

[0166] The method in the embodiments of the present application realizes the full - process digital management of potential hazard identification, task assignment, handling tracking, and result acceptance by constructing an automated closed - loop management mechanism. In this way, after a potential hazard is discovered, a report is automatically generated and the responsible person is notified through various reminder methods. At the same time, the handling process is recorded and the over - due reminder function is supported. Through the comprehensive recording and filing of the closed - loop management, the efficiency of potential hazard handling and the traceability of management are ensured, the operation and maintenance responsibilities are clarified, and a complete technical guarantee is provided for potential hazard investigation and treatment.

[0167] The method in the embodiments of the present application constructs an integrated solution from data collection to closed - loop handling of potential hazards. Data collection provides basic data support, data processing ensures the quality of input data, multi - modal analysis realizes a comprehensive assessment of the internal state and external characteristics of the gas drainer, automatic inspection reduces manual intervention during the inspection process, closed - loop management realizes the whole - process tracking and control of potential hazard treatment, and active learning enables the system to have the ability of adaptive optimization. Each step progresses layer by layer in function and forms a closed - loop in data flow, which can not only ensure real - time performance and accuracy, but also realize the long - term management of the system through learning and optimization, greatly improving the efficiency of potential hazard investigation of the gas drainer, while reducing the safety risks during the inspection process, and providing strong support for the intelligent safety management of iron and steel enterprises.

[0168] In the hidden danger investigation method of the embodiments of the present application, remote automatic video inspection is initiated, and no operation by the inspector is required. Three-dimensional AI intelligent inspection can be used to check key points outside and inside the gas drainer. The inspection standards for each point will be automatically lit to remind, highlighting the key points of inspection. During the automatic inspection process, if hidden dangers are found, real-time alarms will be issued, and inspection problem records will be automatically recorded. After the inspection is completed, closed-loop management of inspection abnormal problems can be carried out. The AI intelligent inspection of the gas drainer has changed the conventional inspection work method, improved the inspection efficiency, enhanced the inspection effect, and reduced the risks of inspection personnel and equipment.

[0169] During the inspection process, if problems are found, the problem pictures will be automatically captured and saved. The remote inspector fills in the abnormal description and specifies the disposal deadline. Click on "Unprocessed Inspection Abnormalities" at the top of the screen to view the detailed information of inspection abnormal problems. Click the confirmation button, indicating that the relevant responsible person has started to handle the problem. After the handling is completed, the relevant responsible personnel can fill in the relevant information to complete the acceptance, thus forming a closed loop. If the disposal and acceptance are not completed by the approaching deadline, the platform will pop up a "Unprocessed Inspection Abnormalities" prompt box for alarm reminder. Remind the relevant responsible personnel to pay attention and hurry up to handle it. You can also click "Extend the Deadline" to appropriately extend the disposal time. The above operation process will be recorded in the inspection ledger, and the staff can export it at any time for easy viewing of the relevant information of the abnormality.

[0170] This method realizes AI intelligent inspection and hidden danger investigation of the operating status of the gas drainer, discovers problems with the drainer in the equipment area in a timely manner, issues alarms immediately, and relevant post personnel can view and handle them, and take effective measures in time to avoid accidents such as gas accidents. It meets the requirements of remote intelligent inspection and hidden danger investigation and abnormal alarm of the drainer, lays a solid technical foundation for cost reduction and efficiency improvement. At the same time, it reduces personnel's contact with the dangerous gas area, avoids accidents, reduces equipment failures, prevents gas leakage, and provides strong support for unattended centralized control of the gas drainer enclosure area and intelligent factory collaborative management and control. Intelligent inspection is more efficient and effective. Shorter inspection time: It takes 60 - 90 seconds to complete the inspection of one drainer, reducing manpower, and there is a substantial improvement compared with manual inspection; better inspection and investigation effect. The inspection points are dynamically lit for reminder, with rules to follow and laws to abide by, eliminating missed inspections; more comprehensive data monitoring and more detailed data. 24-hour real-time monitoring, abnormal alarms for the gas drainer equipment and its auxiliary devices, more comprehensively controlling the status of the gas drainer, and predicting fault risks according to the data curve. Less safety risk: Reducing personnel going to the site, minimizing risks such as poisoning, asphyxiation, slipping, falling, vehicle injuries, and high-altitude falls; clearer operation and maintenance responsibilities. Inspection records, problem handling, overdue alarms, closed-loop management, and the original data are traceable.

[0171] Based on the same inventive concept, the embodiments of the present invention also provide a gas drainer hidden danger investigation device. Figure 3The following is a structural block diagram of a hidden danger detection device for a gas drainer provided by an embodiment of the present invention. As Figure 3 shown, the device 300 includes an acquisition module 301, a first determination module 302, a second determination module 303, a third determination module 304, and a fourth determination module 305.

[0172] The acquisition module 301 is configured to acquire an image and operating status data of the gas drainer, and the operating status data includes at least one of temperature, pressure, liquid level, and carbon monoxide concentration;

[0173] The first determination module 302 is configured to determine a first health score of the gas drainer according to the image;

[0174] The second determination module 303 is configured to determine a second health score of the gas drainer according to the operating status data;

[0175] The third determination module 304 is configured to determine a comprehensive health score of the gas drainer according to the first health score and the second health score;

[0176] The fourth determination module 305 is configured to determine a hidden danger level of the gas drainer according to the comprehensive health score.

[0177] Optionally, the first determination module 302 is further configured to:

[0178] Input the image into a pre-built neural network model to obtain a first health score output by the neural network model;

[0179] Wherein, the neural network model is trained by an image set of the gas drainer, and the image set includes normal operation images and abnormal operation images.

[0180] Optionally, the second determination module 303 is further configured to:

[0181] Input the operating status data into a pre-built memory network model to obtain a second health score output by the memory network model;

[0182] Wherein, the memory network model is trained by an operating status data set of the gas drainer, and the operating status data set includes normal operating status data and abnormal operating status data.

[0183] Optionally, the third determination module 304 is further configured to:

[0184] Obtain a first weight of the first health score and a second weight of the second health score;

[0185] Calculate a first product of the first health score and the first weight, and a second product of the second health score and the second weight;

[0186] Calculate the sum of the first product and the second product to obtain the comprehensive health score of the gas drainer.

[0187] Optionally, the fourth determination module 305 is further configured to:

[0188] If the comprehensive health score is less than or equal to a preset first threshold, determine that the potential hazard level of the gas drainer is a high level;

[0189] If the comprehensive health score is greater than the first threshold and less than a preset second threshold, the potential hazard level of the gas drainer is a medium level;

[0190] If the comprehensive health score is greater than or equal to the second threshold, the potential hazard level of the gas drainer is a low level.

[0191] Optionally, the device 300 further includes an adjustment module for:

[0192] Obtain the false alarm rate of potential hazards of the gas drainer;

[0193] Adjust at least one of the first threshold and the second threshold according to the false alarm rate of potential hazards.

[0194] Optionally, there are multiple gas drainers, and the device 300 further includes an optimization module for:

[0195] Determine the confidence level of the operation status data of each gas drainer;

[0196] Add the operation status data with a confidence level greater than a preset confidence level threshold to the operation status data set to obtain a new operation status data set;

[0197] Optimize the memory network model using the new operation status data set.

[0198] It can be understood that for the device provided in the above embodiments, only the above-mentioned division of each functional module is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0199] An embodiment of the present invention further provides an electronic device, which may include a processor and a memory, and the processor and the memory may be communicatively connected to each other through a bus or other means.

[0200] The processor may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application, or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., or a combination of the above types of chips.

[0201] The memory may include a mass storage for data or instructions. By way of example and not limitation, the memory may include a Hard Disk Drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. In a suitable case, the memory may include removable or non-removable (or fixed) media. In a suitable case, the memory may be internal or external to the electronic device. In a particular embodiment, the memory may be a non-volatile solid-state memory.

[0202] In one example, the memory may be a Read Only Memory (ROM). In one example, the ROM may be a mask-programmed ROM, a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), an Electrically Rewritable ROM (EAROM), or a flash memory, or a combination of two or more of these.

[0203] The processor reads and executes the computer program instructions stored in the memory to implement any one of the gas drainer hidden danger investigation methods in the above embodiments.

[0204] In one example, the electronic device may further include a communication interface and a bus. Among them, the processor, the memory, and the communication interface are connected through the bus and complete communication with each other. The communication interface is mainly used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present application. In a suitable case, the bus may include one or more buses.

[0205] In addition, in combination with the method for detecting potential hazards of the gas drainer in the above embodiments, the embodiments of the present invention can provide a computer-readable storage medium to implement. Computer program instructions are stored on the computer-readable storage medium; when the computer program instructions are executed by a processor, any one of the methods for detecting potential hazards of the gas drainer in the above embodiments is implemented.

[0206] Those skilled in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory, a Hard Disk Drive (HDD), or a Solid-State Drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0207] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages:

[0208] A method, device, equipment, and medium for detecting potential hazards of a gas drainer provided by an embodiment of the present invention obtain an image and operating state data of the gas drainer to understand the external state and internal operating state of the gas drainer. The operating state data at least includes one of temperature, pressure, liquid level, and carbon monoxide concentration; according to the image, a first health score of the gas drainer is determined to evaluate the external health condition; according to the operating state data, a second health score of the gas drainer is determined to evaluate the internal operating state; according to the first health score and the second health score, a comprehensive health score of the gas drainer is determined to understand the comprehensive health condition of the gas drainer; according to the comprehensive health score, the potential hazard level of the gas drainer is determined to detect the potential hazard situation of the gas drainer. This method automatically obtains the comprehensive health score of the gas drainer objectively and accurately through comprehensive analysis of the image and operating state data of the gas drainer without manual participation, so that it can be determined whether there are potential hazards in the gas drainer according to the comprehensive health score.

[0209] In the specification provided here, a large number of specific details are described. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail so as not to obscure the understanding of this specification.

[0210] Similarly, it should be understood that, for the purpose of streamlining the present disclosure and aiding in the understanding of one or more of the various inventive aspects, in the foregoing description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, the inventive aspects lie in less than all the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate embodiment of the present invention.

[0211] It should be noted that the above embodiments are illustrative of the present invention rather than limiting the present invention, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim listing several means, several of these means can be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. does not denote any order. These words can be interpreted as names.

Claims

1. A method for troubleshooting hidden dangers of a gas drainer, characterized in that: The method comprises: Acquire an image and operating status data of the gas drainer, wherein the operating status data includes at least one of temperature, pressure, liquid level and carbon monoxide concentration; determining a first health score of the gas drainer based on the image; determining a second health score of the gas drainer based on the operating status data; determining a comprehensive health score of the gas drainer according to the first health score and the second health score; The hidden danger level of the gas drainer is determined according to the comprehensive health score.

2. The method for troubleshooting gas drainer hazards according to claim 1, characterized in that: Determining a first health score of the gas drainer according to the image includes: Inputting the image into a pre-built neural network model to obtain the first health score output by the neural network model; The neural network model is trained by an image set of the gas drainer, and the image set includes normal operation images and abnormal operation images.

3. The method for troubleshooting gas drainer hazards according to claim 1, characterized in that: Determining a second health score of the gas drainer according to the operating status data includes: Inputting the operating status data into a pre-built memory network model to obtain the second health score output by the memory network model; The memory network model is trained by an operating status data set of the gas drainer, and the operating status data set includes normal operating status data and abnormal operating status data.

4. The method for troubleshooting gas drainer hazards according to claim 1, characterized in that: Determining the comprehensive health score of the gas drainer according to the first health score and the second health score includes: Obtaining a first weight of the first health score and a second weight of the second health score; Calculate a first product of the first health score and the first weight, and a second product of the second health score and the second weight; The sum of the first product and the second product is calculated to obtain a comprehensive health score of the gas drainer.

5. The method for troubleshooting gas drainer hazards according to claim 1, characterized in that: Determining the hidden danger level of the gas drainer according to the comprehensive health score includes: If the comprehensive health score is less than or equal to a preset first threshold, determining that the hidden danger level of the gas drainer is a high level; If the comprehensive health score is greater than the first threshold and less than the preset second threshold, the hidden danger level of the gas drainer is medium; If the comprehensive health score is greater than or equal to the second threshold, the hidden danger level of the gas drainer is a low level.

6. The method for troubleshooting gas drainer hazards according to claim 5, characterized in that: The method further comprises: Obtaining the false alarm rate of hidden dangers of the gas drainer; At least one of the first threshold and the second threshold is adjusted according to the hidden danger false alarm rate.

7. The method for troubleshooting gas drainer hazards according to claim 3, characterized in that: The gas drainer comprises a plurality of gas drainers, and the method further comprises: determining a confidence level of the operating status data of each of the gas drainers; Adding the operating status data whose confidence level is greater than a preset confidence level threshold to the operating status data set to obtain a new operating status data set; The memory network model is optimized using the new operating status data set.

8. A gas drainer hidden danger detection device, characterized in that: The device comprises: An acquisition module, used to acquire an image and operating status data of the gas drainer, wherein the operating status data includes at least one of temperature, pressure, liquid level and carbon monoxide concentration; a first determination module, configured to determine a first health score of the gas drainer according to the image; a second determination module, configured to determine a second health score of the gas drainer according to the operating status data; a third determination module, configured to determine a comprehensive health score of the gas drainer according to the first health score and the second health score; The fourth determination module is used to determine the hidden danger level of the gas drainer according to the comprehensive health score.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 7.

Images