Welding arc detection method and detection device

Abstract

The invention relates to the technical field of electric power system detection, in particular to an electric arc detection method and device for welding, and the method comprises the steps: collecting electrical information in a welding process, carrying out the feature frequency band recognition of the electrical information, generating frequency band recognition information, and carrying out the detection of the electric arc based on the frequency band recognition information. The method comprises the following steps: analyzing the energy ratio of electrical information, obtaining working state data, inputting the working state data into a pre-constructed arc mode recognition model, obtaining an arc fault detection result, rapidly detecting arc faults, improving the response speed, accurately extracting key features, improving the fault detection accuracy, autonomously analyzing the model, reducing human intervention, and improving the fault detection accuracy. Faults are corrected in time, welding stability and quality are ensured, equipment is prevented from being damaged by the faults, the service life is prolonged, system automatic operation is achieved, professional skill requirements are reduced, and whether the electric power system has arc faults or not can be accurately and efficiently judged.

Description

Technical Field

[0001] This application relates to the technical field of power system testing, and in particular to a method and device for detecting welding arc. Background Technology

[0002] Ensuring welding quality and safe operation of welding equipment are crucial during welding. However, the instability of the electric arc during welding can lead to welding defects or even equipment malfunctions. To improve welding quality and reduce equipment failures, it is necessary to monitor and detect the electric arc in real time. Traditional arc detection methods mainly rely on manual observation or simple electrical parameter measurements. These methods are not only inefficient but also inaccurate. Especially when welding is performed on power systems, the frequency bands of the electrical operating parameters are constantly changing. Using fixed identification standards to monitor power systems carries the risk of misjudgment.

[0003] Application content Therefore, it is necessary to provide a welding arc detection method and device to address the above-mentioned technical problems, which can accurately and efficiently determine whether an arc fault has occurred in the power system.

[0004] In a first aspect, this application provides a method for detecting welding arcs, the method comprising: Electrical information of the power system during the welding process is collected, and the characteristic frequency bands of the electrical information are identified to generate frequency band identification information of the electrical information; Based on the frequency band identification information, the electrical information is analyzed in a corresponding form to obtain the power system's operating status data; The operating status data is input into a pre-built arc pattern recognition model to obtain arc fault detection results.

[0005] Secondly, this application also provides a welding arc detection device for implementing the welding arc detection method described in any one of the first aspects, comprising: The frequency band identification module is used to collect electrical information of the power system during the welding process and identify the characteristic frequency bands of the electrical information to generate frequency band identification information of the electrical information; The status analysis module is used to perform energy percentage analysis on the electrical information based on the frequency band identification information to obtain the operating status data of the power system. The fault detection module is used to input the working status data into a pre-built arc pattern recognition model to obtain arc fault detection results.

[0006] The aforementioned arc detection method for welding collects electrical information during the welding process, identifies the characteristic frequency bands of the electrical information, generates frequency band identification information, analyzes the energy proportion of the electrical information based on the frequency band identification information, obtains working status data, inputs the working status data into a pre-constructed arc pattern recognition model, and obtains arc fault detection results. This method can quickly detect arc faults, improve response speed, accurately extract key features, improve fault detection accuracy, and the model can perform autonomous analysis, reducing human intervention, correcting faults in a timely manner, ensuring welding stability and quality, preventing equipment damage from faults, extending service life, and automating system operation, reducing professional skill requirements. It can accurately and efficiently determine whether an arc fault has occurred in the power system. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the steps of a welding arc detection method in one embodiment; Figure 2 This is a schematic diagram of the arc detection device for welding in one embodiment. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0009] The arc detection method for welding provided in this application can be applied to, for example... Figure 1 In the application environment shown: S1: Collect electrical information of the power system during the welding process, and identify the characteristic frequency bands of the electrical information to generate frequency band identification information of the electrical information; S2: Based on the frequency band identification information, perform energy percentage analysis on the electrical information in a corresponding form to obtain the power system's operating status data; S3: Input the working status data into the pre-built arc pattern recognition model to obtain the arc fault detection result.

[0010] Specifically, in step S1 of the embodiments provided in this application, during the welding process, sensors or measuring devices (such as current transformers, voltage transformers, etc.) are used to continuously collect electrical information of the power system, including current, voltage, power, frequency, etc. During the electric welding process, the electrical information of the power system can reflect the welding status and the formation and disappearance of the electric arc, obtain comprehensive and real-time electrical data, and provide basic data for subsequent analysis.

[0011] More specifically, time-frequency domain analysis methods (such as short-time Fourier transform (STFT), wavelet transform (WT)) are used to analyze electrical information and generate time-frequency spectrum diagrams. Arc faults have different characteristics in the time and frequency domains. These characteristics can be extracted through time-frequency domain analysis to obtain the distribution of electrical information in time and frequency, which facilitates the identification of arc characteristic frequency bands.

[0012] More specifically, feature data, including amplitude, frequency, and phase, are extracted from the time-spectrum graph. Feature data can reflect the changing patterns and dynamics of electrical information and is the basis for frequency band identification and fault detection. Extracting representative feature data lays the foundation for subsequent frequency band identification and pattern recognition.

[0013] More specifically, based on the extracted feature data, the characteristic frequency bands of electrical information are identified, and frequency band identification information is generated. It should be noted that the function of frequency band identification information is to divide electrical information into several time zones and assign corresponding characteristic frequency bands to each time zone. In other words, the characteristic frequency bands corresponding to the power system in each time zone are different.

[0014] More specifically, during welding operations in a power system, the welding work will be performed in different characteristic frequency bands at different times. If the same identification standard is used to detect faults in the electrical information of the power system, false detections may occur. Therefore, the electrical information is first divided into frequency bands, and then corresponding feature identification is performed separately to provide a basis for the detection of arc faults.

[0015] Specifically, in step S2 of the embodiment provided in this application, the energy in each frequency band is calculated using frequency band identification information. The specific method can be to calculate the power spectral density (PSD) in each frequency band using spectrum analysis techniques such as Fourier transform (FFT). The energy in the frequency band reflects the energy distribution of the electrical signal at different frequencies and is the basis for energy proportion analysis. Obtaining the energy of each frequency band provides data support for subsequent energy proportion calculation.

[0016] More specifically, the energy of each frequency band is normalized so that the sum of the energy of each frequency band is 1 (or 100%). Normalization can eliminate the difference in the absolute value of energy of different frequency bands, making the energy proportion more comparable. The normalized frequency band energy proportion data is obtained, which is convenient for subsequent analysis and comparison. Based on the normalized frequency band energy data, the proportion of energy of each frequency band to the total energy is calculated to form energy proportion data. The energy proportion can reflect the distribution of energy of each frequency band in the overall electrical information and is an important indicator for assessing the working status of the power system. Detailed energy proportion data provides a basis for subsequent status assessment.

[0017] More specifically, characteristic parameters reflecting the operating status of the power system are extracted from the energy proportion data, such as the energy proportion of the main frequency band and the energy change trend of the frequency band. These characteristic parameters can reflect the changing patterns of the operating status of the power system and are an important basis for condition monitoring and fault detection. Extracting key characteristic parameters that can reflect the operating status of the power system lays the foundation for condition assessment and fault detection.

[0018] More specifically, by using state characteristic parameters, combined with preset thresholds and standards, the operating status of the power system is evaluated to determine whether there are any abnormalities. By evaluating state characteristic parameters, abnormal states of the power system can be detected in a timely manner, preventing potential faults from occurring. Real-time evaluation of the operating status of the power system allows for timely detection and early warning of abnormalities, thereby improving the safety and reliability of system operation.

[0019] Specifically, in step S3 of the embodiment provided in this application, feature parameters are extracted from the preprocessed working state data. These parameters may include frequency features, time-domain features, and statistical features. Feature extraction can extract key information from the original data, reduce data dimensionality, and improve the recognition efficiency of the model, thereby obtaining feature parameters that can represent the working state and providing input for the pattern recognition model.

[0020] More specifically, the extracted feature parameters are input into a pre-built arc pattern recognition model. The arc pattern recognition module can be a model based on a recognition algorithm, that is, obtaining judgment criteria through data collection and analysis, and setting up an algorithm around this judgment rule to perform the fault detection task. This model can be a machine learning (such as SVM, decision tree) or deep learning (such as CNN, RNN) model. The model uses the pattern recognition model to analyze and classify the input feature parameters to detect arc faults. The model can perform calculations and judgments based on the input feature parameters and output arc fault detection results. The model infers based on the input feature parameters and outputs fault detection results, including whether an arc fault exists and the fault type. The model inference is based on the patterns learned during training to make judgments and obtain fault detection results, thus obtaining specific arc fault detection results that indicate the current working status and fault information of the power system.

[0021] More specifically, the arc fault detection results output by the model are verified and compared with actual conditions to evaluate the accuracy and reliability of the model. Verification can ensure the credibility of the model output results, reduce false alarms and false negatives, improve the accuracy and reliability of fault detection results, and ensure the safety of system operation.

[0022] This application provides a method for detecting welding arc, which has the following advantages: The aforementioned arc detection method for welding collects electrical information during the welding process, identifies the characteristic frequency bands of the electrical information, generates frequency band identification information, analyzes the energy proportion of the electrical information based on the frequency band identification information, obtains working status data, inputs the working status data into a pre-constructed arc pattern recognition model, and obtains arc fault detection results. This method can quickly detect arc faults, improve response speed, accurately extract key features, improve fault detection accuracy, and the model can perform autonomous analysis, reducing human intervention, correcting faults in a timely manner, ensuring welding stability and quality, preventing equipment damage from faults, extending service life, and automating system operation, reducing professional skill requirements. It can accurately and efficiently determine whether an arc fault has occurred in the power system.

[0023] In one embodiment, the step of collecting electrical information of the power system during the welding process and identifying the characteristic frequency bands of the electrical information to generate frequency band identification information of the electrical information includes: S11: Continuously acquire electrical signals from the power system performing welding work, and combine the electrical signals at each moment in a time sequence to obtain the electrical information of the power system during the welding process; S12: Perform time-frequency domain analysis on the electrical information to obtain a time-frequency spectrum, and extract time-domain and frequency-domain features based on the time-frequency spectrum to obtain the time-frequency domain feature data of the electrical signal; S13: Identify the characteristic frequency bands of the electrical information based on the time-frequency domain feature data to obtain the frequency band identification information of the electrical information.

[0024] Specifically, sensors and data acquisition equipment are used to collect electrical signals of the power system in real time during the welding process. The collected electrical signals are arranged in chronological order to form a continuous electrical information data stream. The electrical signals in the welding process have dynamic change characteristics, which require real-time acquisition to capture instantaneous changes. The chronological arrangement ensures the integrity and continuity of the data, making subsequent analysis more accurate and obtaining complete and continuous electrical information data, thus laying the foundation for subsequent analysis.

[0025] More specifically, methods such as Short Time Fourier Transform (STFT) and Wavelet Transform are used to convert the time-domain signal into a time-frequency domain signal, generating a time-spectrum graph. The time-spectrum graph shows the distribution of the signal at different times and frequencies. Time-frequency domain analysis can simultaneously display the time and frequency characteristics of the signal, providing more comprehensive information. The time-spectrum graph is an important foundation for feature extraction, revealing the inherent laws of the signal. The obtained time-spectrum graph shows the detailed characteristics of the electrical signal during the welding process.

[0026] More specifically, time-domain and frequency-domain features are extracted based on the time-spectrum graph. Time-domain features, such as amplitude, phase, energy, and peak value, are extracted from the time-spectrum graph, while frequency-domain features, such as frequency components, band energy, and spectral entropy, are extracted. The time-domain and frequency-domain features together describe the time-varying and frequency characteristics of the signal. Feature extraction transforms high-dimensional time-frequency data into low-dimensional feature data, which is convenient for subsequent processing, resulting in time-frequency domain feature data containing rich information, providing a basis for frequency band identification.

[0027] More specifically, based on time-frequency domain feature data, the electrical information is identified by characteristic frequency bands. Machine learning algorithms (such as SVM, random forest, neural network, etc.) or rule-based methods are used to analyze the time-frequency domain feature data to identify characteristic frequency bands. The identified characteristic frequency bands and their corresponding time-domain information are recorded. Machine learning algorithms can automatically identify complex patterns, improving efficiency and accuracy. Characteristic frequency band identification can discover specific patterns in electrical signals, which is helpful for welding quality monitoring and fault detection. The frequency band identification information of electrical information is obtained, the characteristic frequency bands of electrical signals are clarified, and the accuracy and reliability of welding process monitoring are improved.

[0028] In one embodiment, the step of identifying the characteristic frequency bands of the electrical information based on the time-frequency domain feature data to obtain the frequency band identification information of the electrical information includes: S31: Time-align the time-frequency domain feature data with the electrical information to predict the characteristic frequency bands of the electrical information at each time using the time-frequency domain feature data, so as to generate prediction information at each time. S32: Perform correlation analysis on the prediction information at adjacent times, and combine the prediction information at each time based on the results of the correlation analysis to generate the frequency band identification information of the electrical information.

[0029] Specifically, the time-frequency domain feature data is synchronized with the original electrical information to ensure that the feature data at each moment matches the corresponding electrical signal. If there is a time offset or sampling rate difference, necessary time correction and interpolation are performed. Time alignment ensures that the feature data and electrical information are accurately matched, avoids analysis errors, maintains timing consistency, and ensures that subsequent analysis is based on accurate time information. This yields time-aligned time-frequency domain feature data and electrical information, laying the foundation for subsequent feature band prediction.

[0030] More specifically, the electrical information at each time step is predicted using the time-frequency domain feature data. An appropriate prediction model (such as a deep learning model or support vector machine) is constructed or selected. The time-frequency domain feature data is input, and the electrical information at each time step is predicted for the characteristic frequency bands. The prediction results for each time step are output, forming a continuous sequence of prediction information. Using an advanced prediction model can improve the accuracy of characteristic frequency band identification. The prediction information at each time step can reflect the dynamic changes of the electrical signal, which is convenient for real-time monitoring and analysis. The prediction information for the characteristic frequency bands at each time step is obtained, forming a continuous sequence of prediction results.

[0031] More specifically, correlation analysis, smoothing, or dynamic time warping (DTW) are used to perform correlation analysis on the prediction information at adjacent times, identify and mark the prediction information of frequency bands with abnormal changes or discontinuities. Correlation analysis can smooth the prediction information, reduce noise and abnormal fluctuations. By analyzing the correlation between adjacent times, the continuous change trend of electrical signals can be captured, and smoothed prediction information can be obtained, thereby improving the continuity and reliability of the prediction results.

[0032] More specifically, by combining the results of correlation analysis with the prediction information at each time point, the final frequency band identification information is generated. Continuous characteristic frequency bands are marked and their corresponding time periods are recorded. By combining the results of correlation analysis, characteristic frequency bands can be identified and marked more accurately. The comprehensive combination improves the accuracy and reliability of the identification results, generates complete and accurate electrical information frequency band identification information, and clarifies the temporal changes and duration of characteristic frequency bands.

[0033] In one embodiment, a correlation analysis is performed on the prediction information at adjacent time points, and the prediction information at each time point is combined based on the results of the correlation analysis to generate the frequency band identification information of the electrical information. The specific steps include: Correlation analysis is performed on the prediction information at adjacent time points to obtain the consistency characteristics of the prediction information at adjacent time points; Based on the consistency characteristics of adjacent time points, the prediction information of each time point is aggregated and analyzed to aggregate the time points whose consistency characteristics meet the preset criteria into an interval, and to generate the characteristic frequency band to which each interval belongs for the consistency characteristics corresponding to each interval, so as to obtain the frequency band identification information. A tendency analysis of the predicted information of the boundary positions of adjacent intervals is performed to redistribute the predicted information of the boundary positions based on the analysis results.

[0034] Specifically, the consistency characteristics of the prediction information at adjacent time points are calculated using methods such as correlation coefficient, mutual information, and Euclidean distance. For each pair of prediction information at adjacent time points, their feature similarity is calculated to form a consistency feature matrix. By calculating the consistency features, the similarity of the prediction information at adjacent time points can be effectively measured, thereby identifying continuous feature frequency bands. The consistency features help smooth the prediction information and reduce abrupt changes caused by noise. The resulting consistency feature matrix of the prediction information at adjacent time points provides a data foundation for subsequent aggregation analysis.

[0035] More specifically, cluster analysis is performed based on the consistency characteristics of adjacent time points. Based on preset clustering criteria (such as consistency characteristic thresholds), adjacent time points that meet the conditions are clustered into an interval. Clustering algorithms (such as K-means, DBSCAN, etc.) are used to cluster the consistency characteristics, forming multiple intervals. The prediction information within each interval is comprehensively analyzed to extract the characteristic frequency band to which the interval belongs, and corresponding frequency band identification information is generated for each interval. Through cluster analysis, continuous and consistent time information can be grouped into an interval, improving the accuracy and stability of frequency band identification. The generated interval characteristic frequency bands can effectively describe the characteristics of the interval, facilitating subsequent analysis and application. Multiple intervals with consistent characteristics and their corresponding characteristic frequency band information are obtained, generating accurate frequency band identification information that reflects the characteristic frequency band changes of electrical information.

[0036] More specifically, a tendency analysis is performed on the predicted information at the boundary positions of adjacent intervals. For each interval's boundary position (i.e., the predicted information at the interval boundary), a tendency analysis is conducted to determine which adjacent interval it is more inclined towards. Trend analysis, transition probability, and other methods are used to analyze the tendency of the predicted information at the boundary positions. Based on the results of the tendency analysis, the predicted information at the boundary positions is redistributed to ensure it belongs to the appropriate interval. The interval boundaries are updated to ensure consistency and continuity within the intervals. Tendency analysis can optimize interval boundaries, reduce misallocation, and improve the continuity and consistency of information within intervals. By redistributing the predicted information at the boundary positions, interval transitions can be smoothed, abrupt changes and discontinuities can be reduced, resulting in optimized interval boundaries. This ensures the continuity and consistency of information within each interval. The improved interval division and frequency band identification information are more accurate and reflect more realistic electrical signal characteristics.

[0037] In one embodiment, the step of performing energy percentage analysis on the electrical information based on the frequency band identification information to obtain the power system's operating status data includes: S21: Divide the electrical information into several intervals, and each interval has a corresponding characteristic frequency band; S22: Retrieve the pre-deployed specified identification form according to the characteristic frequency band corresponding to each interval, and analyze the basic value component and fluctuation component of the electrical information in the interval based on the specified identification form, and analyze the total energy ratio of the fluctuation component in each interval to obtain the fluctuation energy ratio index of each interval. S23: Analyze the overall working status of the electrical information based on the fluctuation energy ratio index of each interval to obtain the working status data of the power system.

[0038] Specifically, based on frequency band identification information, a time axis is configured for the electrical information, arranging the electrical information in chronological order. The characteristic frequency band corresponding to each time point is identified. Based on the time axis and frequency band identification information, the electrical information is divided into several intervals, each interval corresponding to a characteristic frequency band. This ensures that the electrical information within each interval has similar frequency band characteristics. Configuring a time axis for the electrical information enables the data to have temporal continuity and sequence, facilitating subsequent analysis. Dividing the electrical information into intervals helps to gather information with the same characteristic frequency bands, facilitating unified analysis and processing. This generates time-series electrical information data, providing a foundation for subsequent analysis and yielding multiple characteristic intervals, each interval corresponding to specific frequency band characteristics.

[0039] More specifically, a specified identification method is retrieved and the base value and fluctuation components are analyzed. Based on the characteristic frequency bands corresponding to each interval, a pre-deployed specified identification method is retrieved. The specified identification method may include Fourier transform, wavelet transform, etc., to analyze the frequency components of electrical information. The specified identification method is applied to analyze the electrical information in each interval, extract the base value component (DC component) and fluctuation component (AC component), calculate the spectral characteristics of the base value component and fluctuation component in each interval, and obtain the energy value of each component. Through the specified identification method, the frequency components of electrical information can be analyzed in detail, the base value and fluctuation components can be distinguished, and the energy of the base value and fluctuation components can be analyzed. This helps to understand the energy distribution characteristics of electrical information. The base value component and fluctuation component of electrical information are successfully separated, the spectral characteristics of each component are obtained, and the energy values ​​of the base value and fluctuation components in each frequency band are extracted, laying the foundation for energy proportion analysis.

[0040] More specifically, the total energy proportion analysis of the fluctuation components calculates the total energy proportion of the fluctuation components within each interval, i.e., the proportion of the fluctuation component energy to the total energy of the entire interval. Through normalization, the energy proportion is converted into a fluctuation energy proportion index. The fluctuation component energy within each interval is statistically and normally processed to obtain the fluctuation energy proportion index for each interval. The fluctuation component energy proportion analysis can reveal the energy characteristics of the fluctuation components in electrical information and reflect the dynamic characteristics of the system. By calculating the fluctuation energy proportion index, the energy characteristics of different intervals can be standardized, facilitating comparison and analysis. The fluctuation energy proportion index quantifies the energy characteristics of the fluctuation components within each interval, providing a standardized energy characteristic description for subsequent working state analysis.

[0041] More specifically, the overall operating status is analyzed and operating status data is generated. The fluctuation energy proportion index of each interval is assigned to the corresponding interval on the time axis, forming an energy proportion map on the time axis. This reflects the energy distribution characteristics of electrical information over the entire time range. Based on the energy proportion map on the time axis, the overall operating status of electrical information is analyzed. Combined with other auxiliary data (such as voltage, current, etc.), the operating status of the power system is further analyzed. The analysis results are organized into operating status data, including the fluctuation energy proportion index at each moment and other relevant indicators. The operating status data is output in the form of charts, reports, etc. By assigning the fluctuation energy proportion index to the time axis, the energy characteristics of electrical information can be examined from a global perspective, and the operating status of the system can be identified. Combined with the fluctuation energy proportion index and other auxiliary data, a comprehensive operating status analysis of the power system can be performed, potential problems or anomalies can be discovered, and an energy proportion map on the time axis is generated. This intuitively displays the changes in the energy distribution of electrical information, obtains detailed operating status data, and provides a comprehensive view of the power system operation, facilitating monitoring and maintenance.

[0042] In one embodiment, the steps for parsing the overall operating status of electrical information include: S331: Analyze the time characteristics of each interval on the time axis to assign corresponding time characteristic data to each fluctuation energy proportion index; S332: Combining the characteristic frequency bands corresponding to each interval and the time characteristic data, assign corresponding signal recognition tasks to each interval; S333: Based on the signal identification task, identify the abnormality of electrical signals in the specified interval, and generate abnormal characteristic data based on the identification results; S334: Assign the time characteristic data and the abnormal characteristic data to the corresponding fluctuation energy ratio index, and combine the fluctuation energy ratio indices together as working status data.

[0043] Specifically, the time characteristics of each interval on the time axis are analyzed, including calculating the duration, start time, and end time of each interval, and extracting the time characteristic data of each interval, such as time length and periodicity. Analyzing the time characteristic data can correlate the data of each interval with time, which helps to understand the changes of electrical information in the time dimension. Extracting the time characteristic data helps to further analyze the behavioral characteristics of each interval, obtain the time characteristic data of each interval, and provide a time dimension reference for subsequent analysis. The electrical information of each interval is closely related to the time axis, which facilitates time series analysis.

[0044] More specifically, each fluctuation energy proportion index is assigned corresponding time characteristic data. The parsed time characteristic data is then assigned to each fluctuation energy proportion index to ensure that each fluctuation energy proportion index contains time characteristic data for the corresponding interval. Assigning time characteristic data makes the fluctuation energy proportion index more complete, including information in the time dimension. With time characteristic data, the fluctuation energy proportion index can be comprehensively analyzed in the time dimension. The fluctuation energy proportion index contains time characteristic data, forming a complete dataset. The fluctuation energy proportion index obtains information in the time dimension, which is convenient for time series analysis.

[0045] More specifically, corresponding signal recognition tasks are assigned to each interval. Based on the characteristic frequency bands and time characteristics data of each interval, corresponding signal recognition tasks are assigned. Signal recognition tasks may include spectrum analysis, pattern recognition, etc. By assigning targeted signal recognition tasks according to the characteristic frequency bands and time characteristics of different intervals, the accuracy of recognition can be improved. The overall recognition task is decomposed into each interval, making the recognition work more targeted and effective. Each interval receives a targeted signal recognition task, which improves the accuracy and efficiency of recognition. The signal recognition tasks are clear, which facilitates subsequent execution and management.

[0046] More specifically, anomaly identification of electrical signals is performed within a specified interval. Based on the assigned signal identification task, anomaly identification of electrical signals is carried out within the specified interval. Anomaly detection algorithms (such as statistical anomaly detection, machine learning algorithms, etc.) are used to identify abnormal features in the electrical signals. Based on the identification results, abnormal characteristic data is generated, including anomaly type, abnormal time period, and anomaly degree. Anomaly identification of electrical signals can promptly detect and locate abnormal situations in the system. The generation of abnormal characteristic data helps to further analyze and handle abnormal situations. Identifying abnormal situations in electrical signals and generating abnormal characteristic data helps to locate the interval and specific characteristics of the problem, providing a basis for fault handling.

[0047] More specifically, time-related data and abnormal characteristic data are assigned to the corresponding fluctuation energy proportion index. This ensures that each fluctuation energy proportion index contains time-related and abnormal characteristic data for its corresponding interval. The fluctuation energy proportion indices assigned with time-related and abnormal characteristic data are integrated to form complete working status data. By associating time-related and abnormal characteristic data, the fluctuation energy proportion index becomes more comprehensive and facilitates integrated analysis. The integrated fluctuation energy proportion index data is more complete, covering multi-dimensional information such as time, energy, and abnormalities. Each fluctuation energy proportion index contains time-related and abnormal characteristic data, forming comprehensive working status data. The working status data has multi-dimensional information, enabling more comprehensive and in-depth analysis.

[0048] In one embodiment, the time characteristics of each interval on the time axis are analyzed to assign corresponding time characteristic data to each fluctuation energy proportion index. The steps include: The duration of each interval on the time axis is analyzed to obtain the duration characteristic data of each interval; The time position occupied by each interval on the time axis is located to obtain the positioning characteristic data of each interval; Based on the positioning characteristic data of each interval, the frequency band distribution pattern of each interval with the same characteristic frequency band is analyzed to obtain the frequency band distribution characteristics of various characteristic frequency bands, and then converted into corresponding distribution characteristic data to be assigned to the corresponding intervals. The continuous characteristic data, location characteristic data, and distribution characteristic data of each interval are combined and compressed to generate time characteristic data.

[0049] Specifically, the duration of each interval on the time axis is calculated. The duration of each interval is calculated according to the start and end times of the interval. The purpose of calculating the duration is to obtain the time length of each interval, which is convenient for subsequent feature analysis. The duration is the basic time characteristic data and the basis for further analysis. Obtaining the duration characteristic data of each interval provides a basis for subsequent analysis. Obtaining the time length of each interval makes it easier to understand the time distribution of the interval.

[0050] More specifically, the time position of each interval on the timeline is located, the specific position of each interval on the timeline is determined, the start and end times of each interval are determined, and the time position data is recorded. Time position location is to determine the specific position of each interval on the timeline, which is convenient for subsequent time series analysis. By using time position, the characteristics of different intervals can be distinguished to avoid confusion, the positioning characteristic data of each interval can be obtained, its position on the timeline can be clarified, and time distribution information can be provided, which helps to analyze the time relationship between intervals.

[0051] More specifically, based on the positioning characteristic data of each interval, the frequency band distribution pattern of intervals with the same characteristic frequency bands is analyzed. This analysis examines the distribution pattern of various characteristic frequency bands on the time axis, extracts frequency band distribution features, converts these features into corresponding distribution characteristic data, and assigns this data to the corresponding intervals. Analyzing the distribution pattern of characteristic frequency bands allows us to understand the distribution pattern of each frequency band on the time axis. Assigning the distribution characteristic data to each interval ensures that each interval contains frequency band distribution information, resulting in frequency band distribution characteristic data that reflects the distribution of various characteristic frequency bands. The inclusion of frequency band distribution characteristics in each interval facilitates further frequency band analysis.

[0052] More specifically, the continuous characteristic data, location characteristic data, and distribution characteristic data of each interval are combined and compressed. This process integrates all characteristic data to form a complete dataset. The combined information is then compressed to reduce data redundancy and optimize data storage and transmission efficiency. Appropriate compression algorithms (such as data encoding and clustering) are used for information compression. Data combination ensures that all characteristic data are integrated to form complete interval time characteristic data. Information compression reduces data redundancy, optimizes data storage and transmission efficiency, improves system performance, and generates complete time characteristic data containing continuous, location, and distribution characteristic information. Through information compression, efficient data storage and transmission are achieved, enhancing system performance.

[0053] In one embodiment, the process of constructing an arc pattern recognition model includes: S301: Deploy electrical systems in both normal and fault states to perform welding work, and collect electrical information and observation data of each electrical system during the welding work; S302: Identify the characteristic frequency bands of the electrical information, and based on the identification results, analyze the energy ratio, time characteristics, and abnormal data of the intervals corresponding to each characteristic frequency band of the electrical information to obtain the electrical characteristics of the corresponding electrical information; S303: Evaluate the working status of the electrical system based on the observation data, and encode the evaluation results to obtain the performance characteristics corresponding to the observation data; S304: Analyze the correlation between the electrical characteristics and the performance characteristics to obtain the data relationship between the electrical system in normal and fault states; S305: Adjust the parameters of the pre-deployed recognition algorithm according to the data connection form to construct an arc pattern recognition model.

[0054] Specifically, the electrical system is deployed under both normal and fault conditions to simulate actual welding work. During the welding process, electrical information (such as voltage, current, and power) and observation data (such as welding quality and the number of sparks) are collected to obtain the operating data of the electrical system under both normal and fault conditions, ensuring the comprehensiveness and diversity of the data.

[0055] More specifically, the collected electrical information is subjected to spectrum analysis to identify characteristic frequency bands. Energy proportion analysis is performed on the interval of each characteristic frequency band to determine the proportion of each frequency band in the total energy. Time characteristic analysis is performed to determine the distribution of characteristic frequency bands on the time axis. Abnormal data is detected and analyzed to identify abnormal patterns and behaviors. Key features in the electrical information are extracted to lay the foundation for subsequent feature coding and model construction.

[0056] More specifically, the working status of the electrical system is evaluated based on observation data (such as welding quality, number of sparks, etc.) to determine whether the system is in a normal or faulty state. The evaluation results are encoded to generate performance features corresponding to the observation data, and the working status evaluation results are converted into encoded features that can be used for model training.

[0057] More specifically, we conduct correlation analysis on electrical characteristics and performance characteristics to find the correspondence between electrical information and operating status. We use statistical analysis, correlation calculation and other methods to determine the characteristic patterns under normal and fault conditions, understand the characteristic differences of electrical systems under different states, and find the characteristic relationship forms under normal and fault conditions.

[0058] More specifically, based on the obtained data connection format, the parameters of the pre-deployed recognition algorithm (such as machine learning model, neural network, etc.) are adjusted, and the input features, weights, learning rate and other parameters of the model are adjusted so that the model can accurately identify the arc pattern and build a high-precision arc pattern recognition model that can effectively distinguish between normal and fault states.

[0059] More specifically, using the adjusted parameters and collected feature data, an arc pattern recognition model is trained. The accuracy of the model is verified through methods such as cross-validation and test set evaluation, ensuring that the constructed arc pattern recognition model has high accuracy and reliability in practical applications.

[0060] In one embodiment, when constructing an arc pattern recognition model based on the data connection format, a prediction verification mechanism is further included, the prediction verification mechanism comprising: S3001: Deploy a prediction algorithm based on the arc pattern recognition model; the prediction algorithm is used to perform a correction analysis on the basic detection results of the arc pattern recognition model to obtain several corrected detection results, and based on the basic detection results and each of the corrected detection results, predict future information for the existing collected electrical information to generate prediction verification information. S3002: Deploy a verification algorithm based on the prediction algorithm; the verification algorithm is used to match the electrical information collected in the future according to the prediction verification information, obtain the confidence level of the basic detection result and each of the corrected detection results, and judge the basic detection result and each of the corrected detection results according to the confidence level, so as to correct the basic detection result.

[0061] Specifically, the electric arc pattern recognition model is used to analyze the currently collected electrical information to obtain basic detection results, namely the preliminary judgment of the electric arc state (normal or fault). Error analysis is performed on the basic detection results. By adjusting the model parameters or applying different analysis methods, several corrective detection results are generated. These results are corrections to the basic detection results to different degrees. Through multiple corrective detection results, the comprehensiveness and reliability of the judgment of the electric arc state are improved.

[0062] More specifically, based on the basic test results and the results of each correction test, the existing collected electrical information is used to predict future information, predict the trend of electrical information change over a period of time, generate prediction verification information, and apply time series prediction models (such as LSTM, ARIMA, etc.) to predict the future trend of electrical information in advance, providing data support for verifying the basic test results and correction test results.

[0063] More specifically, based on the predicted verification information, it is matched with the electrical information actually collected in the future. Based on the matching results, the confidence level of the basic test results and each corrective test result is calculated. The confidence level reflects the degree of agreement between the predicted results and the actual data. The reliability of the test results is evaluated by the confidence level, and the most credible arc state judgment is selected.

[0064] More specifically, based on the confidence level of each detection result, the basic detection results and the corrected detection results are comprehensively judged to obtain the final arc fault detection result. If the arc state cannot be effectively judged based on the confidence level, an operation adjustment instruction is generated for the power system to switch to a specific operating state, ensuring that, under uncertain circumstances, data can be further collected and the model can be verified by adjusting the system operating state.

[0065] More specifically, when the power system switches to a specific operating state, electrical information is re-collected, the detection standards for that specific operating state are invoked, and the newly collected electrical information is subjected to feature recognition to obtain feature recognition results. Under the specific operating state, the accuracy of the detection results is improved through the new feature recognition standards.

[0066] More specifically, the feature recognition results are combined with the previously calculated confidence level for comprehensive judgment. Based on the comprehensive result, the final arc fault detection result is determined. Through multiple verifications and comprehensive analysis, the accuracy and reliability of the final arc fault detection result are ensured.

[0067] In one embodiment, such as Figure 2 As shown, a welding arc detection device is provided for implementing the welding arc detection method according to any one of the first aspects, comprising: The frequency band identification module is used to collect electrical information of the power system during the welding process and identify the characteristic frequency bands of the electrical information to generate frequency band identification information of the electrical information; The status analysis module is used to perform energy percentage analysis on the electrical information based on the frequency band identification information to obtain the operating status data of the power system. The fault detection module is used to input the working status data into a pre-built arc pattern recognition model to obtain arc fault detection results.

[0068] In this embodiment, the specific implementation of each module in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0069] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0071] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0072] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting welding arc, characterized in that, include: Electrical information of the power system during the welding process is collected, and the characteristic frequency bands of the electrical information are identified to generate frequency band identification information of the electrical information; Based on the frequency band identification information, the electrical information is analyzed in a corresponding form to obtain the power system's operating status data; The operating status data is input into a pre-built arc pattern recognition model to obtain arc fault detection results.

2. The arc detection method for welding as described in claim 1, characterized in that, The steps of collecting electrical information from a power system during the welding process and identifying the characteristic frequency bands of the electrical information to generate frequency band identification information for the electrical information include: The electrical signals of the power system performing welding work are continuously collected, and the electrical signals at each moment are combined in a time sequence manner to obtain the electrical information of the power system during the welding process. The electrical information is analyzed in the time-frequency domain to obtain a time-spectrum diagram, and time-domain and frequency-domain features are extracted based on the time-spectrum diagram to obtain the time-frequency domain feature data of the electrical signal. Based on the time-frequency domain feature data, the electrical information is identified by characteristic frequency bands to obtain frequency band identification information of the electrical information.

3. The arc detection method for welding as described in claim 2, characterized in that, The step of identifying the characteristic frequency bands of the electrical information based on the time-frequency domain feature data to obtain the frequency band identification information of the electrical information includes: The time-frequency domain feature data is time-aligned with the electrical information to predict the characteristic frequency bands of the electrical information at each time using the time-frequency domain feature data, thereby generating prediction information for each time. A correlation analysis is performed on the prediction information at adjacent time points, and the prediction information at each time point is combined based on the results of the correlation analysis to generate the frequency band identification information of the electrical information.

4. The arc detection method for welding as described in claim 1, characterized in that, The steps for performing energy percentage analysis on the electrical information based on the frequency band identification information to obtain the power system's operating status data include: The electrical information is divided into several intervals, and each interval has a corresponding characteristic frequency band; According to the characteristic frequency bands corresponding to each interval, the pre-deployed specified identification form is retrieved, and the electrical information in the interval is analyzed based on the specified identification form. The ratio of the fluctuation component to the total energy of the interval is analyzed for each interval to obtain the fluctuation energy ratio index of each interval. The overall operating status of the electrical information is analyzed based on the fluctuation energy ratio index of each interval to obtain the operating status data of the power system.

5. The arc detection method for welding as described in claim 4, characterized in that, The steps for analyzing the overall operating status of electrical information include: The time characteristics of each interval are analyzed to assign corresponding time characteristic data to each fluctuation energy proportion index. By combining the characteristic frequency bands corresponding to each interval with the time characteristic data, a corresponding signal recognition task is assigned to each interval; According to the signal recognition task, the abnormality of electrical signals in the specified interval is identified, and abnormal characteristic data is generated based on the identification results; The time characteristic data and the abnormal characteristic data are assigned to the corresponding fluctuation energy ratio index, and the fluctuation energy ratio indexes are combined to form the working status data.

6. The arc detection method for welding as described in claim 1, characterized in that, The process of constructing the arc pattern recognition model includes: Deploy electrical systems in both normal and fault states to perform welding work, and collect electrical information and observation data of each electrical system during the welding work; The electrical information is identified by characteristic frequency bands, and based on the identification results, the energy ratio, time characteristics, and abnormal data of each characteristic frequency band interval corresponding to the electrical information are analyzed to obtain the electrical characteristics of the corresponding electrical information. The electrical system's operating status is evaluated based on the observed data, and the evaluation results are feature-encoded to obtain the performance characteristics corresponding to the observed data. The correlation between the electrical characteristics and the performance characteristics is analyzed to obtain the data relationship between the electrical system in normal and fault states; The parameters of the pre-deployed recognition algorithm are adjusted according to the data connection format to construct an arc pattern recognition model.

7. The arc detection method for welding as described in claim 6, characterized in that, When constructing the arc pattern recognition model based on the aforementioned data relationship, a prediction verification mechanism is also included, which includes: A prediction algorithm is deployed on the basis of the arc pattern recognition model. The prediction algorithm is used to perform detection error correction analysis on the basic detection results of the arc pattern recognition model to obtain several corrected detection results. Based on the basic detection results and each of the corrected detection results, the algorithm predicts future information of the existing electrical information to generate prediction verification information. A verification algorithm is deployed based on the prediction algorithm; the verification algorithm is used to match the electrical information collected in the future according to the prediction verification information, to obtain the confidence level of the basic detection result and each of the corrected detection results, and to judge the basic detection result and each of the corrected detection results according to the confidence level, so as to correct the basic detection result.

8. A welding arc detection device, characterized in that, A welding arc detection method according to any one of claims 1-7 includes: The frequency band identification module is used to collect electrical information of the power system during the welding process and identify the characteristic frequency bands of the electrical information to generate frequency band identification information of the electrical information; The status analysis module is used to perform energy percentage analysis on the electrical information based on the frequency band identification information to obtain the operating status data of the power system. The fault detection module is used to input the working status data into a pre-built arc pattern recognition model to obtain arc fault detection results.

Images