Welding fault detection method and device and computer readable storage medium

Abstract

The invention discloses a welding fault detection method and device and a computer readable storage medium, and relates to the technical field of welding, the method comprises the following steps: preprocessing at least one kind of sensor data, and determining standardized data corresponding to each kind of sensor data; performing linear dimension reduction operation on the standardized data to determine dimension-reduced data; training a machine learning model by using the dimension-reduced data, mapping the dimension-reduced data to a low-dimensional space, and determining a low-dimensional data graph; and according to the low-dimensional data graph, data point distribution is checked, abnormal points are identified, and a fault detection result is determined. According to the welding fault detection method provided by the invention, the efficiency and accuracy of welding fault detection are effectively improved through data preprocessing, standardization and dimension reduction in combination with the machine learning model.

Description

Technical Field

[0001] The present application relates to the field of welding technology, and in particular to a welding fault detection method, device and computer-readable storage medium. Background Art

[0002] Welded structures are widely used in various industrial fields, such as aerospace, automobile manufacturing, construction, petrochemicals, etc. These welded structures bear important functions and safety responsibilities, so their quality is directly related to the performance and reliability of the product. During the welding process, due to the influence of various factors such as materials, processes, and equipment, various welding defects may occur, such as cracks, pores, slag inclusions, lack of fusion, and lack of penetration. These defects will seriously affect the performance and service life of the welded structure, and may even cause structural failure and safety accidents.

[0003] At present, welding fault detection mostly adopts manual detection methods, which are carried out through manual identification and perception.

[0004] However, the manual welding fault detection process requires high proficiency and experience of workers, the efficiency of manual detection is not high enough, there is no unified standard in the manual detection process, it is difficult to quantify indicators, and the detection accuracy is low.

[0005] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The main purpose of the present application is to provide a welding fault detection method, device and computer-readable storage medium, aiming to solve the technical problems of low efficiency and accuracy of welding fault detection.

[0007] To achieve the above objectives, the present application proposes a welding fault detection method, which comprises:

[0008] Preprocessing at least one sensor data to determine standardized data corresponding to each sensor data;

[0009] Performing a linear dimensionality reduction operation on the standardized data to determine dimensionality reduction data;

[0010] Using the reduced-dimensional data to train a machine learning model, mapping the reduced-dimensional data to a low-dimensional space, and determining a low-dimensional data graph;

[0011] According to the low-dimensional data graph, the data point distribution is checked, the abnormal points are identified, and the fault detection result is determined.

[0012] In one embodiment, the step of preprocessing at least one sensor data to determine the standardized data corresponding to each sensor data includes:

[0013] Checking the data integrity of each of the sensor data, processing missing values, and determining complete data;

[0014] Remove noise from each of the complete data by using a filter to determine denoised data;

[0015] Synchronize the time of each type of denoised data to determine synchronized data;

[0016] Based on each of the synchronized data, normalized data is determined.

[0017] In one embodiment, before the step of preprocessing at least one sensor data to determine the standardized data corresponding to each sensor data, the step includes:

[0018] Install at least one sensor at a key position of the welding machine;

[0019] The sensor data is recorded in real time and stored.

[0020] In one embodiment, the step of performing a linear dimensionality reduction operation on the standardized data to determine the dimensionality reduced data comprises:

[0021] Calculating a covariance matrix based on the standardized data;

[0022] Perform eigenvalue decomposition according to the covariance matrix to determine eigenvalues ​​and eigenvectors;

[0023] Taking the eigenvector corresponding to the larger eigenvalue among the eigenvectors as the principal component;

[0024] The normalized data is projected onto the principal components to determine the dimensionally reduced data.

[0025] In one embodiment, the step of using the reduced-dimensional data to train a machine learning model, mapping the reduced-dimensional data to a low-dimensional space, and determining a low-dimensional data graph comprises:

[0026] Setting parameters of the machine learning model;

[0027] Inputting the reduced-dimensional data into the machine learning model, mapping the reduced-dimensional data into a low-dimensional space, and determining low-dimensional data;

[0028] The low-dimensional data is visualized to determine a low-dimensional data graph.

[0029] In one embodiment, the step of checking the data point distribution, identifying abnormal points, and determining the fault detection result according to the low-dimensional data graph includes:

[0030] Based on the low-dimensional data graph, examining data point distribution by cluster analysis;

[0031] Setting a threshold for fault detection, and marking points in the data points that exceed the threshold as abnormal points;

[0032] According to the abnormal point, the fault detection result is determined by the alarm system.

[0033] In one embodiment, the step of determining the fault detection result through an alarm system according to the abnormal point includes:

[0034] Establish an alarm system, and when an abnormal point is detected, send out an alarm signal and record the abnormal data corresponding to the abnormal point;

[0035] Based on the abnormal data, a fault detection result is determined.

[0036] In one embodiment, after the step of checking the data point distribution, identifying the abnormal points, and determining the fault detection result according to the low-dimensional data graph, the following steps are included:

[0037] Verify the accuracy of fault detection results through the actual welding status;

[0038] The parameters of the machine learning model are adjusted to improve the welding fault detection method.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a welding fault detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the welding fault detection method described above.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the welding fault detection method described above are implemented.

[0041] One or more technical solutions proposed in this application have at least the following technical effects:

[0042] The present application provides a welding fault detection method, which preprocesses at least one sensor data to determine the standardized data corresponding to each sensor data; performs linear dimensionality reduction operation on the standardized data to determine the reduced dimensionality data; uses the reduced dimensionality data to train a machine learning model, maps the reduced dimensionality data to a low-dimensional space, and determines a low-dimensional data graph; based on the low-dimensional data graph, checks the data point distribution, identifies abnormal points, and determines the fault detection result. The welding fault detection method proposed in the present application effectively improves the efficiency and accuracy of welding fault detection through data preprocessing, standardization and dimensionality reduction, combined with a machine learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0045] Figure 1 A schematic diagram of a flow chart provided for the first embodiment of the welding fault detection method of the present application;

[0046] Figure 2 This is a flow chart of steps S210-S240 in the second embodiment of the welding fault detection method of the present application;

[0047] Figure 3 This is a flow chart of steps S310-S340 in Embodiment 3 of the welding fault detection method of this application;

[0048] Figure 4 This is a flow chart of steps S410-S430 in the fourth embodiment of the welding fault detection method of the present application;

[0049] Figure 5 This is a flow chart of steps S510-S530 in Embodiment 5 of the welding fault detection method of the present application;

[0050] Figure 6 Schematic diagram of the device structure of the hardware operating environment involved in the dynamic correction method of the end of the robot arm in the embodiment of the present application.

[0051] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0053] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0054] The main solution of the embodiment of the present application is: preprocess at least one sensor data to determine the standardized data corresponding to each sensor data; perform linear dimensionality reduction operation on the standardized data to determine the reduced dimensionality data; use the reduced dimensionality data to train a machine learning model, map the reduced dimensionality data to a low-dimensional space, and determine a low-dimensional data graph; based on the low-dimensional data graph, check the data point distribution, identify abnormal points, and determine the fault detection result.

[0055] At present, welding fault detection mostly adopts manual detection methods, which are manually identified and perceived. The manual welding fault detection process requires high proficiency and experience of workers, the efficiency of manual detection is not high enough, there is no unified standard in the manual detection process, it is difficult to quantify indicators, and the detection accuracy is low.

[0056] The present application provides a solution, which ensures the accuracy and consistency of the data by preprocessing at least one sensor data, and then determines the standardized data corresponding to each sensor data, performs a linear dimensionality reduction operation on the standardized data, generates reduced dimensionality data, uses these reduced dimensionality data to train a machine learning model, maps the reduced dimensionality data to a low-dimensional space through the model, and generates a low-dimensional data graph that is easy to observe and analyze. Finally, according to the distribution of data points in the low-dimensional data graph, anomalies are accurately identified, thereby determining the fault detection results and improving the efficiency and accuracy of welding fault detection.

[0057] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a welding fault detector, etc. The following takes the welding fault detector as an example to illustrate this embodiment and the following embodiments.

[0058] Embodiment 1

[0059] Embodiment 1 of the present application discloses a welding fault detection method, referring to Figure 1 , the welding fault detection method comprises:

[0060] Step S110: pre-process at least one sensor data to determine standardized data corresponding to each sensor data.

[0061] In this embodiment, the sensor data is the measurement value or observation value generated by the sensor, including temperature, pressure, vibration, welding speed and other data. The preprocessing is to clean and reduce the noise of the data, and the standardized data is the data converted into a standard scale.

[0062] As an optional implementation, outliers or missing values ​​due to sensor failure, transmission errors, etc. are deleted or repaired, filters are used to reduce noise caused by the environment or the sensor itself, time series data is smoothed, the data set is traversed to find the maximum and minimum values ​​of the data, and the data is standardized to a standard scale using a standardization formula. Among them, filters include moving average filters, Kalman filters, etc.

[0063] Optionally, before step S110, the following steps are included:

[0064] Step S111, installing at least one sensor at a key position of the welding machine.

[0065] In this embodiment, the welding machine is a device used for welding, which melts the contact surfaces of two or more workpieces and combines them into a whole by heating and / or pressurizing.

[0066] As an optional implementation, according to the structure and function of the welding machine, the key positions where sensors need to be installed are determined, and according to the parameters to be monitored, suitable sensor types are selected and installed. The monitored parameters include temperature, pressure, current, voltage, etc.

[0067] Step S112, recording sensor data in real time and storing the sensor data.

[0068] In this embodiment, real-time recording refers to the process of capturing, recording or saving data immediately when the data is generated.

[0069] As an optional implementation, the sensor sends data to a data acquisition system or controller through its interface, and the data acquisition system or controller periodically or continuously reads the data sent by the sensor, and writes the sensor data stored in the memory or buffer to a long-term storage medium periodically or on demand. Among them, the sensor interface includes analog signals, digital signals, wireless transmission, etc., and the long-term storage medium includes hard disks, flash memory, databases, cloud storage, etc.

[0070] Step S120, performing a linear dimensionality reduction operation on the standardized data to determine dimensionality reduced data.

[0071] In this embodiment, linear dimensionality reduction is a data preprocessing technology used to project high-dimensional data into a low-dimensional space while retaining the original structure and information of the data as much as possible. The reduced dimensionality data is a low-dimensional representation extracted from the original high-dimensional data through dimensionality reduction technology. Dimensionality reduction technology includes principal component analysis PCA, t-SNE, UMAP, etc.

[0072] As an optional implementation, the covariance matrix of the standardized data is calculated, and the covariance matrix is ​​decomposed to obtain the eigenvalues ​​and corresponding eigenvectors. According to the size of the eigenvalues, the eigenvectors corresponding to the larger eigenvalues ​​are selected as new basis vectors, and the standardized data are projected onto the new basis vectors to obtain the reduced-dimensional data.

[0073] Exemplarily, based on the standardized data, the covariance matrix between each feature is calculated, and the covariance matrix is ​​eigenvalue decomposition is performed to obtain a set of eigenvalues ​​and corresponding eigenvectors. Based on the size of the eigenvalues, the eigenvectors corresponding to the first k (k represents the target dimension to be reduced) largest eigenvalues ​​are selected as principal components. The selected principal components (eigenvectors) are used to construct a projection matrix, and the standardized data are projected onto this matrix to obtain the reduced-dimensional data.

[0074] Step S130, use the reduced-dimensional data to train a machine learning model, map the reduced-dimensional data to a low-dimensional space, and determine a low-dimensional data graph.

[0075] In this embodiment, the machine learning model is a data-based algorithm or method for learning from data and making predictions or decisions, including decision trees, random forests, support vector machines, neural networks, etc.

[0076] As an optional implementation, select a suitable machine learning model, use the reduced dimension data as input features, use the reduced dimension data to train the machine learning model, select an appropriate visualization tool, select an appropriate graphic type according to the dimension and nature of the reduced dimension data, use the visualization tool to draw the reduced dimension data graphics, and determine the low-dimensional data graphics. Among them, the visualization tools include matplotlib, seaborn, plotly, etc., and the graphic types include scatter plots, heat maps, etc.

[0077] Step S140, checking the data point distribution according to the low-dimensional data graph, identifying abnormal points, and determining the fault detection result.

[0078] In this embodiment, the data point distribution refers to the arrangement and clustering of data points in a visualization graph, and an outlier point refers to a point in a data set that is significantly different from other data points.

[0079] As an optional implementation, the arrangement and aggregation of data points in the low-dimensional data graph are observed, and possible abnormal points are preliminarily identified based on the distribution of the data points, and the correlation factors are comprehensively considered to determine whether there are fault detection results. The correlation factors include the number, location, distribution of abnormal points, and the correlation with the actual business scenario.

[0080] In this embodiment, at least one sensor data is preprocessed to ensure the accuracy and consistency of the data, and then the standardized data corresponding to each sensor data is determined, and a linear dimensionality reduction operation is performed on the standardized data to generate reduced dimensionality data. The machine learning model is trained using the reduced dimensionality data, and the reduced dimensionality data is mapped to a low-dimensional space through the model to generate a low-dimensional data graph that is easy to observe and analyze. Finally, based on the distribution of data points in the low-dimensional data graph, abnormal points are accurately identified to determine the fault detection results, thereby improving the efficiency and accuracy of welding fault detection.

[0081] Based on the first embodiment, the second embodiment of the present application proposes a welding fault detection method, referring to Figure 2 , step S110 includes:

[0082] Step S210, checking the data integrity of each type of sensor data, processing missing values, and determining complete data.

[0083] In this embodiment, data integrity refers to the state in which data is complete, not damaged or lost during the data collection process, and missing values ​​are data points that are missing or unavailable in the data set.

[0084] As an optional implementation, data is read from the sensor, the integrity of the data is checked, missing values ​​in the data set are identified, and the missing values ​​are deleted, filled or replaced to determine complete data.

[0085] For example, read the raw data from the sensor or storage medium, check whether the data is recorded in the expected format and frequency, identify which data points in the data set are missing, and if there are few data points with missing values, or the missing values ​​have little impact on data analysis, directly delete the data rows or columns containing the missing values, find the data points before and after the missing values, use the linear interpolation formula to estimate the size of the missing values, and fill them into the data set. After interpolation and filling, a complete data set is obtained.

[0086] Step S220: removing noise from each of the complete data by using a filter to determine denoised data.

[0087] In this embodiment, the filter is a device or algorithm for removing unnecessary components or noise from a signal, and filter types include low-pass filter, high-pass filter, band-pass filter, band-stop filter, digital filter, etc.

[0088] As an optional implementation, a suitable filter type is selected, and the filter parameters are determined according to the characteristics of the sensor and the data quality, and the complete data is input into the filter to perform denoising and determine the denoised data. The filter parameters include cutoff frequency, window size, number of iterations, etc.

[0089] Exemplarily, a moving average filter is selected to remove noise from the temperature data. The window size is determined based on the noise level of the temperature data and the required degree of smoothness. The entire temperature data sequence is traversed. For each data point, the average value of the data point and the two data points before and after it (a total of 5 data points) is calculated. The calculated average value is used as the denoised data point, and all denoised data points are combined into a new data sequence, namely, denoised data.

[0090] Step S230, synchronizing the time of each type of denoised data to determine the synchronized data.

[0091] In this embodiment, the synchronization data time is aligned through a timestamp, and the timestamp is used to record the time when each data point is collected.

[0092] As an optional implementation, identify timestamps in the denoised data, convert all timestamps into a uniform format, sort the data by timestamp, and create a merged data set containing data from all data sources to determine the synchronized data.

[0093] Step S240: determining standardized data based on each type of synchronization data.

[0094] In this embodiment, standardized data refers to converting raw data according to certain standards or specifications so that it has a unified format, unit and measurement, so as to enable comparison and analysis across sensors, platforms or time.

[0095] As an optional implementation, for each feature in the data set, its mean and standard deviation are calculated, and a standardization formula is applied to each data point in the data set to generate a new data set, namely, standardized data.

[0096] For example, a data set of synchronized data contains two features, temperature and pressure. The original dimensions of these two features are different (temperature may be Celsius and pressure may be Pascal), so their scales may also be different. The mean and standard deviation of the temperature data and pressure data are calculated respectively, and Z-score normalization is applied. For each temperature data point and pressure data point in the data set, their Z-score is calculated to determine the standardized data.

[0097] In this embodiment, by cleaning, denoising and unifying the format of the raw sensor data, redundant information and interference factors in the data can be eliminated, making the subsequent analysis more focused and accurate. By converting the data of different sensors to the same scale, it can be ensured that various data have the same weight and importance in subsequent analysis. This helps the machine learning model better understand and identify patterns in the data, thereby improving the accuracy of fault detection.

[0098] Based on the first embodiment, the third embodiment of the present application proposes a welding fault detection method, referring to Figure 3 , step S120 includes:

[0099] Step S310, calculating a covariance matrix based on the standardized data.

[0100] In this embodiment, the covariance matrix is ​​a square matrix, whose elements are the covariances between the ith and jth random variables. The covariance matrix describes the linear relationship between the random variables, especially the correlation.

[0101] Exemplarily, an n×n zero matrix C is initialized to store the elements of the covariance matrix. For each element C_ij (i=1,2,...,n; j=1,2,...,n) of the covariance matrix, the covariance between the i-th column and the j-th column in Z is calculated, that is, C_ij=Σ_k=1^m(Z_ki-μ_i)*(Z_kj-μ_j) / (m-1), where μ_i and μ_j are the means of the i-th column and the j-th column in Z, respectively (since it has been standardized, μ_i=μ_j=0), and the calculated C_ij is filled into the i-th row and j-th column of the matrix C. The final matrix C is the covariance matrix of the standardized data.

[0102] Step S320: Perform eigenvalue decomposition according to the covariance matrix to determine eigenvalues ​​and eigenvectors.

[0103] In this embodiment, eigenvalue decomposition is a method of decomposing a matrix into a set of eigenvectors and eigenvalues. The eigenvalue describes the degree of stretching or compression of the matrix with respect to the eigenvector. The eigenvector is a non-zero vector that makes the transformation of the matrix a simple telescopic transformation.

[0104] As an optional implementation, the covariance matrix is ​​subjected to eigenvalue decomposition using a function provided by a numerical computing library, to obtain eigenvalues ​​and corresponding eigenvectors. The numerical computing library includes NumPy, MATLAB, and the like.

[0105] Step S330: taking the eigenvector corresponding to the larger eigenvalue among the eigenvectors as the principal component.

[0106] In this embodiment, principal component is a concept in principal component analysis (PCA), which refers to the direction with the largest change in a data set. The direction corresponding to the eigenvector with a larger eigenvalue is the principal component, which represents the main change direction in the data set.

[0107] As an optional implementation, the eigenvalues ​​and eigenvectors are sorted according to the size of the eigenvalues ​​to ensure that the eigenvalues ​​are arranged from large to small, and the order of the eigenvectors is adjusted accordingly, and the first k eigenvectors with larger eigenvalues ​​are selected as principal components. Among them, k is a hyperparameter and needs to be selected according to actual conditions.

[0108] Step S340, projecting the standardized data onto the principal component to determine the dimension-reduced data.

[0109] In this embodiment, projection refers to the process of mapping a point, line or figure from one space to another space. In this embodiment, projection refers to mapping standardized data onto principal components to obtain data after dimensionality reduction.

[0110] As an optional implementation, the standardized data is projected onto the selected principal components to obtain dimensionally reduced data.

[0111] Exemplarily, a dot product operation is performed on each sample in the high-dimensional standardized data and each selected principal component (ie, eigenvector) to obtain a 2-dimensional or 3-dimensional vector, which is the data after dimensionality reduction.

[0112] In this embodiment, high-dimensional data is mapped to low-dimensional space through dimensionality reduction technology, while retaining key information in the data, which not only reduces the computational complexity, but also facilitates intuitive observation and analysis of data distribution and patterns, thereby improving the efficiency of the welding fault detection method.

[0113] Based on the first embodiment, the fourth embodiment of the present application proposes a welding fault detection method, referring to Figure 4 , step S130 includes:

[0114] Step S410, setting parameters of the machine learning model.

[0115] In this embodiment, the machine learning model refers to a model obtained by training with a machine learning algorithm and capable of predicting or classifying input data. In the machine learning model, parameters refer to adjustable variables within the model that are used to determine the behavior of the model.

[0116] As an optional implementation, the parameters of the machine learning model are initialized and a suitable learning rate is selected. The initialization methods include random initialization, constant initialization, normal distribution initialization, etc. The learning rate is an important hyperparameter in machine learning, which determines the step size of each parameter update during the training process.

[0117] Exemplarily, the t-SNE algorithm is used to reduce the dimensionality of a high-dimensional data set for visualization, and the perplexity parameter (determines the number of neighboring points considered when constructing the conditional probability distribution in the high-dimensional space) is set to 30, the learning rate parameter (determines the speed at which the algorithm updates the parameters during the optimization process) is set to 'auto', the n_iter parameter (determines the number of iterations of the optimization process) is set to 1000 iterations, and the early_exaggeration parameter (in the initial stage of optimization, the distance between points will be enlarged to help the algorithm better capture the local structure of the data) is set to 4.

[0118] Step S420: input the reduced-dimensional data into the machine learning model, map the reduced-dimensional data to a low-dimensional space, and determine the low-dimensional data.

[0119] In this embodiment, low-dimensional data is the result of the dimensionality reduction process, which is the representation of the original high-dimensional data in the low-dimensional space. Such data usually has fewer features or dimensions, but can still reflect the main features or structure of the original data.

[0120] As an optional implementation, the reduced-dimensional data is input into a selected machine learning model, the machine learning model is trained using the reduced-dimensional data, the reduced-dimensional data is mapped to a low-dimensional space, and low-dimensional data is determined to reflect the structure of the high-dimensional data.

[0121] Step S430, visualize the low-dimensional data and determine a low-dimensional data graph.

[0122] In this embodiment, visualization is the process of converting data or information into graphics, images or animations to make it easier for people to understand, analyze and interpret patterns and trends in the data. Visualization methods include scatter plots, heat maps, parallel coordinate graphs, etc. Low-dimensional data graphics refer to the results of displaying low-dimensional data in the form of graphics or images, including scatter plots, heat maps, line graphs, etc., which are used to intuitively represent patterns and relationships in the data.

[0123] As an optional implementation, a visualization tool is used to plot the low-dimensional data into a graph, wherein the visualization tool includes Matplotlib, Seaborn, Plotly, Tableau, etc.

[0124] Exemplarily, use the visualization tool Matplotlib to draw a scatter plot, determine a graphic area large enough to display all data points, select a unique color for each handwritten digit category, iterate over the data points of each digit category, draw scatter points in the graphic area, each data point corresponds to a two-dimensional coordinate (obtained by t-SNE dimensionality reduction), and is represented by the color of the corresponding category, add a legend to the graph, the legend should contain each digit category and its corresponding color, add a title at the top or bottom of the graph, and determine the low-dimensional data graph.

[0125] In this embodiment, by using reduced dimensionality data to train a machine learning model, during real-time detection, the model can quickly identify abnormal data points and give accurate fault detection results. The visualized output results make the data distribution and patterns of the welding process intuitively visible. In the subsequent analysis process, the visualization results are displayed in a graphical form, which can save the time of fault detection result analysis and improve the accuracy of fault detection result analysis.

[0126] Based on the first embodiment, the fifth embodiment of the present application proposes a welding fault detection method, referring to Figure 5 , step S140 includes:

[0127] Step S510, based on the low-dimensional data graph, checking the distribution of data points through cluster analysis;

[0128] In this embodiment, cluster analysis is an unsupervised learning method used to group similar data points together.

[0129] As an optional implementation, according to the characteristics of the data and the analysis objectives, a clustering algorithm is selected, and the low-dimensional data is used as input to run the selected clustering algorithm. The algorithm will divide the data points into different clusters according to the similarity between the data points, and superimpose the clustering results on the low-dimensional data graph to show the distribution and clustering of the data points. Among them, clustering algorithms include K-means clustering, hierarchical clustering, DBSCAN (density-based spatial clustering with noise applications), spectral clustering, etc.

[0130] Step S520, setting a threshold for fault detection, and marking the data points exceeding the threshold as abnormal points;

[0131] In this embodiment, the threshold is a boundary value used to distinguish between two or more data of different states or categories. In fault detection, the threshold is usually used to determine whether a data point is considered abnormal. Anomalies refer to data points that are significantly different from most data points. They may be atypical data generated due to various reasons.

[0132] As an optional implementation, a threshold is determined based on the statistical characteristics of the data, each data point in the data set is traversed to check whether its value exceeds the set threshold, and each data point in the data set is analyzed and compared according to the set fault detection threshold to find abnormal points that exceed the threshold. The statistical characteristics of the data include mean, standard deviation, median absolute deviation, etc.

[0133] As another optional implementation, the threshold is determined based on the quartiles, the median and upper and lower quartiles of the data are used to determine the range of outliers, the outlier range is used as the threshold, and data points in the outlier range are marked as outliers.

[0134] Exemplarily, a distance threshold is determined based on the t-SNE visualization result, and if the distance between a data point and the cluster center exceeds a certain threshold, it is marked as an anomaly.

[0135] Step S530: determining a fault detection result through an alarm system according to the abnormal point.

[0136] In this embodiment, the alarm system is a mechanism for monitoring and detecting abnormal conditions and triggering an alarm notification when an abnormal condition is detected.

[0137] As an optional implementation, the triggering conditions of the alarm system are set, the notification method of the alarm system is configured, and the alarm system determines the fault detection result according to predetermined rules and logic.

[0138] Optionally, step S530 includes:

[0139] Step S531, establishing an alarm system, when an abnormal point is detected, sending out an alarm signal and recording abnormal data corresponding to the abnormal point.

[0140] In this embodiment, the alarm signal is a notification or signal sent when the alarm system is triggered, which is used to inform relevant personnel that an abnormal situation has occurred in the system or data.

[0141] As an optional implementation, according to business needs and data characteristics, the alarm system's monitoring rules and trigger conditions are configured, and the alarm system's notification method is configured. When an abnormal point is detected, the alarm system will automatically trigger an alarm signal according to the preset configuration. While sending the alarm signal, the alarm system records the abnormal data related to the abnormal point. The recorded data may include the original value of the abnormal point, timestamp, and related context information (such as device ID, sensor number, etc.).

[0142] Step S532: determining a fault detection result based on the abnormal data.

[0143] In this embodiment, the fault detection result is a conclusion or evaluation result on whether a system or device has a fault based on the analysis and processing of abnormal data. This result may include information such as the type, location, and severity of the fault, which is used to guide subsequent fault detection and repair work.

[0144] As an optional implementation, based on the analysis of abnormal data, the possible fault type is determined, the specific location or component where the fault occurred is determined based on the context information of the abnormal data, and the severity of the fault is evaluated based on the characteristics of the abnormal data and the analysis results.

[0145] Optionally, after step S530, the following steps are included:

[0146] Step S533, verifying the accuracy of the fault detection result through the actual welding state.

[0147] In this embodiment, the actual state of welding is the actual condition during the welding process.

[0148] As an optional implementation, actual status data of the welded parts are collected during the welding process, and the fault detection results are compared and analyzed with the actual welding status data. The detected fault type is compared to see whether it is consistent with the fault type in the actual state, the fault location is compared to see whether it is accurate, and the assessment of the fault severity is compared to see whether it is accurate. Based on the results of the comparative analysis, the accuracy of the fault detection method is evaluated.

[0149] Step S534, adjusting the parameters of the machine learning model to improve the welding fault detection method.

[0150] In this embodiment, model parameters are internal values ​​that need to be adjusted during the training process of the machine learning model. These parameters determine how the model fits and predicts the data.

[0151] As an optional implementation method, if there is a difference between the detection result and the actual state, the cause is analyzed and the parameters are adjusted based on the evaluation results of the model. Methods such as grid search, random search or Bayesian optimization can be used to find the optimal parameter combination. The process of model training, evaluation and parameter adjustment is repeated until the performance of the model reaches a satisfactory level, thereby improving the welding fault detection method.

[0152] In this embodiment, the accuracy of the fault detection results is verified by the actual welding state, and the parameters of the machine learning model are adjusted according to the verification results to improve the welding fault detection method, which can significantly improve the accuracy of welding fault detection.

[0153] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the welding fault detection method of the present application. More simple transformations based on this technical concept are all within the protection scope of the present application.

[0154] The present application provides a welding fault detection device, which includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the welding fault detection method in the above-mentioned embodiment 1.

[0155] Reference below Figure 6 , which shows a schematic diagram of the structure of a welding fault detection device suitable for implementing the embodiment of the present application. The welding fault detection device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The welding fault detection device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0156] like Figure 6As shown, the welding fault detection device may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the xxx device are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the welding fault detection device to communicate with other devices wirelessly or by wire to exchange data. Although the welding fault detection device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0157] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0158] The welding fault detection device provided by the present application adopts the welding fault detection method in the above embodiment, which can solve the technical problem of welding fault detection. Compared with the prior art, the beneficial effects of the welding fault detection device provided by the present application are the same as the beneficial effects of the welding fault detection method provided by the above embodiment, and other technical features in the welding fault detection device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0159] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0160] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0161] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the welding fault detection method in the above-mentioned embodiment.

[0162] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0163] The computer-readable storage medium may be included in the welding fault detection device; or may exist independently without being assembled into the welding fault detection device.

[0164] The computer-readable storage medium carries one or more programs, and when the one or more programs are executed by the welding fault detection device, the welding fault detection device can write computer program codes for performing the operations of the present application in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, through the Internet using an Internet service provider).

[0165] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0166] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0167] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned welding fault detection method, and can solve the technical problem of welding fault detection. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the welding fault detection method provided in the above-mentioned embodiment, and will not be elaborated here.

[0168] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A welding fault detection method, characterized in that: The welding fault detection method comprises: Preprocessing at least one sensor data to determine standardized data corresponding to each sensor data; Performing a linear dimensionality reduction operation on the standardized data to determine dimensionality reduction data; Using the reduced-dimensional data to train a machine learning model, mapping the reduced-dimensional data to a low-dimensional space, and determining a low-dimensional data graph; According to the low-dimensional data graph, the data point distribution is checked, the abnormal points are identified, and the fault detection result is determined.

2. The welding fault detection method according to claim 1, characterized in that: The step of preprocessing at least one sensor data to determine the standardized data corresponding to each sensor data comprises: Checking the data integrity of each of the sensor data, processing missing values, and determining complete data; Remove noise from each of the complete data by using a filter to determine denoised data; Synchronize the time of each type of denoised data to determine synchronized data; Based on each of the synchronized data, normalized data is determined.

3. The welding fault detection method according to claim 1, characterized in that: Before the step of preprocessing at least one sensor data to determine the standardized data corresponding to each sensor data, the method includes: Install at least one sensor at a key position of the welding machine; The sensor data is recorded in real time and stored.

4. The welding fault detection method according to claim 1, characterized in that: The step of performing a linear dimensionality reduction operation on the standardized data to determine the dimensionality reduction data comprises: Calculating a covariance matrix based on the standardized data; Perform eigenvalue decomposition according to the covariance matrix to determine eigenvalues ​​and eigenvectors; Taking the eigenvector corresponding to the larger eigenvalue among the eigenvectors as the principal component; The normalized data is projected onto the principal components to determine the dimensionally reduced data.

5. The welding fault detection method according to claim 1, characterized in that: The step of using the reduced-dimensional data to train a machine learning model, mapping the reduced-dimensional data to a low-dimensional space, and determining a low-dimensional data graph comprises: Setting parameters of the machine learning model; Inputting the reduced-dimensional data into the machine learning model, mapping the reduced-dimensional data into a low-dimensional space, and determining low-dimensional data; The low-dimensional data is visualized to determine a low-dimensional data graph.

6. The welding fault detection method according to claim 1, characterized in that: The step of checking data point distribution, identifying abnormal points, and determining fault detection results according to the low-dimensional data graph comprises: Based on the low-dimensional data graph, examining data point distribution by cluster analysis; Setting a threshold for fault detection, and marking points in the data points that exceed the threshold as abnormal points; According to the abnormal point, the fault detection result is determined by the alarm system.

7. The welding fault detection method according to claim 6, characterized in that: The step of determining the fault detection result through the alarm system according to the abnormal point includes: Establish an alarm system, and when an abnormal point is detected, send out an alarm signal and record the abnormal data corresponding to the abnormal point; Based on the abnormal data, a fault detection result is determined.

8. The welding fault detection method according to claim 1, characterized in that: After the step of checking the data point distribution, identifying the abnormal points, and determining the fault detection result according to the low-dimensional data graph, the method further comprises: Verify the accuracy of fault detection results through the actual welding status; The parameters of the machine learning model are adjusted to improve the welding fault detection method.

9. A welding fault detection device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the welding fault detection method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the welding fault detection method according to any one of claims 1 to 8 are implemented.

Images