Electron beam welding state detection method and system

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a method and system for detecting the state of electron beam welding, the method comprising: obtaining a voltage data sequence in a welding process; interpolating missing data in the voltage data sequence using an HTFE algorithm, calculating a correction value of a prediction interval error factor in the HTFE algorithm when interpolating each missing data, and replacing the corresponding missing data with voltage data at a corresponding moment predicted based on the correction value to obtain a complete data sequence, so as to perform abnormality detection on the complete data sequence, and detecting the welding state based on the abnormality detection result. The present invention can accurately detect the welding state of an electron beam.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more specifically, to an electron beam welding state detection system. Background Art

[0002] Electron beam welding is a high-precision, high-efficiency technology that can weld at high temperatures and in small areas. The detection of welding status is crucial to ensure welding quality and improve welding efficiency. The welding voltage directly affects the focus and heat input of the electron beam. High voltage can increase the welding depth, but may cause the heat-affected zone to be too large, affecting the quality of the weld; low voltage may result in insufficient welding depth and a weak weld. Therefore, by detecting the voltage data, the welding status can be detected. However, during the data transmission process, electromagnetic interference may cause data loss and affect the detection effect. Therefore, before detecting the voltage data, the missing data needs to be interpolated to ensure the integrity of the data.

[0003] In the related technology, a patent application document with publication number CN113536921A discloses an arc welding quality assessment method, device and storage medium, the method comprising: obtaining a welding current signal and a welding voltage signal; obtaining voltage and current distance data from the welding current signal and the welding voltage signal according to a symbolic approximation aggregation algorithm and a distance algorithm; inputting the voltage and current distance data into an assessment model to obtain a welding quality assessment label.

[0004] However, the above scheme does not consider the impact of missing data in welding current signal and welding voltage signal when using training sample set for model training and evaluation. The presence of missing data in the training sample set may cause the model to be unable to fully learn the true correlation between signals, thus affecting the accurate prediction and evaluation of welding process, resulting in inaccurate evaluation of welding quality, and further affecting the accuracy of welding state detection. Summary of the invention

[0005] In order to solve the problem that the welding state cannot be accurately detected due to the existence of missing data, the present invention provides an electron beam welding state detection method and system.

[0006] According to a first aspect of the present invention, there is provided a method for detecting an electron beam welding state, comprising:

[0007] Acquire voltage data sequence during welding;

[0008] The missing data in the voltage data sequence are interpolated using the HTFE algorithm. When interpolating each missing data, the correction value of the prediction interval error factor in the HTFE algorithm is calculated, and the corresponding missing data is replaced by the voltage data at the corresponding moment predicted based on the correction value to obtain a complete data sequence, so as to perform anomaly detection on the complete data sequence, and detect the welding state based on the anomaly detection result;

[0009] The method for obtaining the correction value includes: taking a sequence consisting of several voltage data before any missing data as a reference sequence, calculating the primary correction coefficient of the prediction interval error factor of the missing data : ; , Respectively The missing data in the reference sequence The sample values ​​and predicted values ​​of the data; is the amount of data in the reference sequence; is the number of extreme points in the reference sequence; is the variance of the sampling intervals between adjacent extreme points in the reference sequence; is the absolute value symbol;

[0010] The correlation between the voltage data and the current data collected in the corresponding period of the reference sequence is calculated, and the normalized value of the product of the correlation and the primary correction coefficient is used as the weight to weight the prediction interval error factor of the missing data to obtain the correction value.

[0011] When the present invention uses the HTFE algorithm to interpolate the missing data in the voltage data sequence, it can specifically adjust the prediction interval error factor of each missing data, avoiding the over-simplification caused by using a unified prediction interval error factor, so that the HTFE algorithm can maintain high accuracy in various complex environments, thereby ensuring the accuracy of the complete data sequence obtained after the interpolation is completed, and then based on the complete data sequence with high accuracy, accurate detection of the welding state can be achieved.

[0012] Preferably, the correlation is a comprehensive correlation, and the comprehensive correlation satisfies the following relationship:

[0013] ;

[0014] In the formula, For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; For the The Pearson correlation coefficient of the voltage data and current data collected during the corresponding period of the reference sequence with missing data; , Respectively The kurtosis of the voltage data and the kurtosis of the current data collected during the corresponding period of the reference sequence with missing data; To preset hyper parameters; is the absolute value symbol.

[0015] The present invention calculates the comprehensive correlation in order to evaluate the degree of distortion of the data in the reference sequence, so that the primary correction coefficient of each missing data can be corrected based on the degree of distortion.

[0016] Preferably, the normalized value of the product of the comprehensive correlation and the primary correction coefficient is used as the final correction coefficient of the prediction interval error factor corresponding to the missing data. The final correction coefficient satisfies the following relationship:

[0017] ;

[0018] In the formula, For the The final correction coefficient of the prediction interval error factor for missing data; For the The primary correction coefficient for the prediction interval error factor for missing data; For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; is the normalization function.

[0019] The present invention utilizes comprehensive correlation to correct the primary correction coefficient, which can avoid the influence of distorted data, and thus can accurately correct the prediction interval error factor of each missing data.

[0020] Preferably, after acquiring the voltage data sequence during the welding process, the method further comprises:

[0021] The voltage data sequence is curve fitted using the least square method to obtain a fitting curve of the voltage data sequence.

[0022] The present invention fits the voltage data sequence in order to reduce the difficulty of data analysis.

[0023] Preferably, the method for obtaining the number of extreme value points in the reference sequence includes:

[0024] The difference method is used to identify the extreme points in the reference sequence, and the extreme points in the reference sequence are counted in turn to obtain the number of extreme points in the reference sequence.

[0025] Preferably, performing anomaly detection on the complete data sequence includes:

[0026] The LOF anomaly detection algorithm is used to perform anomaly detection on the complete data sequence.

[0027] Preferably, detecting the welding state based on the abnormal detection result includes:

[0028] When any voltage data in the voltage data sequence is determined to be an abnormal value, the welding state of the electron beam at the time corresponding to the voltage data is determined to be an abnormal state;

[0029] When it is determined that any voltage data in the voltage data sequence is a normal value, it is determined that the welding state of the electron beam at the time corresponding to the voltage data is a normal state.

[0030] The present invention can realize accurate detection of welding status.

[0031] According to a second aspect of the present invention, there is provided an electron beam welding state detection system, the system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the first aspect of the present invention.

[0032] The present invention has the following effects:

[0033] 1. The present invention can accurately interpolate missing data in a voltage data sequence, thereby enabling accurate detection of welding status based on a complete data sequence with high accuracy.

[0034] 2. The interpolation result determined by the HTFE algorithm in the present invention can better adapt to the volatility and abnormal characteristics of the data, and can effectively avoid the adaptive deviation caused by distorted data, thereby ensuring the accuracy of the interpolation result. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0036] Figure 1 It is a schematic diagram of the steps of a method for detecting electron beam welding status according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0038] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0039] Reference Figure 1 , a method for detecting electron beam welding state, comprising steps S1 to S3, specifically as follows:

[0040] S1: Obtain the voltage data sequence during the welding process.

[0041] Specifically, the welding voltage can be collected during the electron beam welding process using a pressure sensor within a certain time range (e.g., 1 hour) and at a certain collection frequency (e.g., 5 times per second) to obtain a voltage data sequence during the welding process. The present invention does not specifically limit the collection time and collection frequency of the welding voltage.

[0042] In an exemplary embodiment of the present invention, after acquiring the voltage data sequence during the welding process, the following steps are further included:

[0043] The voltage data sequence is curve fitted using the least square method to obtain a fitting curve of the voltage data sequence.

[0044] It should be noted that the voltage data sequence is fitted to facilitate the subsequent analysis process. In the subsequent analysis process, the operations performed on the voltage data sequence are all based on the fitting curve of the voltage data sequence.

[0045] Optionally, the voltage data sequence may be fitted using a weighted least squares method, a smoothing spline method, or the like, so as to obtain a fitting curve of the voltage data sequence. This embodiment does not specifically limit the method for determining the fitting curve of the voltage data sequence. It should be noted that the process of fitting a curve to a data sequence using the least squares method is a prior art and is not described in detail in this embodiment.

[0046] S2: Use the HTFE algorithm to interpolate the missing data in the voltage data sequence. When interpolating each missing data, calculate the correction value of the prediction interval error factor in the HTFE algorithm, and replace the corresponding missing data with the voltage data at the corresponding moment predicted based on the correction value to obtain a complete data sequence.

[0047] It should be noted that during the transmission of welding voltage data, strong electromagnetic interference may affect the stability of the signal and cause data loss. The incompleteness of the data will affect the accurate monitoring of the welding voltage. Therefore, before performing abnormality detection on the voltage data sequence, the present invention uses the HTFE algorithm to interpolate the missing data in the voltage data sequence, so as to obtain a complete data sequence.

[0048] It should be further explained that when the HTFE algorithm is used to interpolate the missing data in the voltage data sequence, the algorithm will apply the same prediction interval error factor to each missing data in the voltage data sequence. However, the unified prediction interval error factor will reduce the adaptability and prediction accuracy of the algorithm, thereby affecting the accuracy of the interpolated data. Therefore, the present invention improves the algorithm to achieve accurate adaptation of the prediction interval error factor of each missing data when the HTFE algorithm is used to preset the voltage data at the corresponding moment of the missing data, thereby improving the adaptability and prediction accuracy of the algorithm and achieving accurate interpolation of the missing data.

[0049] The prediction interval error factor is a parameter used in the HTFE algorithm to determine the size of the prediction interval. It should be noted that the prediction interval error factor is a professional term in the HTFE algorithm and is not described in detail in this embodiment.

[0050] Specifically, the correction value of the prediction interval error factor for any missing data can be determined by the following steps:

[0051] Step 1: Take the sequence of several voltage data before any missing data as the reference sequence, and calculate the primary correction coefficient of the prediction interval error factor of the missing data;

[0052] The primary correction coefficient refers to a parameter used to correct the prediction interval error factor of any missing data, determined based on the data characteristics in the reference sequence of any missing data. It should be noted that when the prediction interval error factor of any missing data is corrected using the primary correction coefficient, the initial correction coefficient can be directly multiplied by the prediction interval error factor of the missing data, thereby correcting the prediction interval error factor of the missing data.

[0053] Optionally, when the primary correction coefficient corresponding to any missing data is larger, the error factor of the prediction interval after the missing data is corrected is relatively larger; conversely, when the primary correction coefficient corresponding to any missing data is smaller, the error factor of the prediction interval after the missing data is corrected is relatively smaller.

[0054] Optionally, a sequence consisting of 50 voltage data before any missing data may be used as a reference sequence for the missing data. This embodiment does not specifically limit the amount of data in the reference sequence.

[0055] Specifically, the primary correction coefficient of the prediction interval error factor for any missing data satisfies the following relationship:

[0056] ;

[0057] In the formula, For the The primary correction coefficient for the prediction interval error factor for missing data; , Respectively The missing data in the reference sequence The sampled values ​​and predicted values ​​of the data, wherein the predicted values ​​are determined based on the HTFE algorithm; is the amount of data in the reference sequence. =50; is the number of extreme points in the reference sequence; is the variance of the sampling intervals between adjacent extreme points in the reference sequence; is the absolute value symbol.

[0058] in, Quantified the The prediction error in the reference sequence with missing data reflects the performance of the HTFE algorithm in the historical process; the larger the value, the larger the prediction error of the data before the missing data, and a larger prediction interval error factor needs to be set to increase the prediction interval and accommodate larger error fluctuations. When it is large, the primary correction coefficient of the prediction interval error factor corresponding to the missing data should also be large, thereby increasing the prediction interval error factor corresponding to the missing data.

[0059] Quantified the The fluctuation of the data in the reference sequence with missing data reflects the data changes in the historical process; the larger the value, the more drastic the data changes in the reference sequence with missing data, and a larger prediction interval error factor needs to be set to increase the prediction interval and accommodate larger error fluctuations. When it is large, the primary correction coefficient of the prediction interval error factor corresponding to the missing data should also be large, thereby increasing the prediction interval error factor corresponding to the missing data.

[0060] In an exemplary embodiment of the present invention, the determination of the number of extreme value points in the reference sequence can be achieved by the following steps:

[0061] The difference method is used to identify the extreme points in the reference sequence, and the extreme points in the reference sequence are counted in turn to obtain the number of extreme points in the reference sequence.

[0062] Optionally, the extreme value points in the reference sequence may be determined by using a peak detection method, a maximum and minimum neighborhood method, etc. This embodiment does not specifically limit the method for determining the extreme value points in the data sequence.

[0063] Step 2: Calculate the correlation between the voltage data and the current data collected during the corresponding period of the reference sequence;

[0064] It should be noted that, since strong electromagnetic interference may cause data loss or distortion, if there is distorted data in the reference sequence of any missing data in the voltage data sequence, the accuracy and credibility of the primary correction coefficient corresponding to the missing data will be reduced. In the welding process, voltage and current are usually related, which is specifically manifested as follows: a larger welding current usually causes a larger welding arc, and the length of the arc is proportional to the voltage, so that an increase in welding current will lead to an increase in arc voltage. Therefore, the present invention calculates the correlation between voltage data and current data in the same period to evaluate the degree of distortion of voltage data in the corresponding period, thereby correcting the primary correction coefficient and realizing accurate correction of the prediction interval error factor of any missing data.

[0065] In an exemplary embodiment of the present invention, the correlation between the voltage data and the current data collected during the corresponding period of the reference sequence is a comprehensive correlation. Specifically, the comprehensive correlation between the voltage data and the current data collected during the corresponding period of the reference sequence satisfies the following relationship:

[0066] ;

[0067] In the formula, For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; For the The Pearson correlation coefficient of the voltage data and the current data collected during the corresponding period of the reference sequence with missing data. It should be noted that the method for determining the Pearson correlation coefficient between the data is the existing technology and will not be described in detail in this embodiment; , Respectively The kurtosis of the voltage data and the kurtosis of the current data collected during the corresponding period of the reference sequence with missing data. It should be noted that the difference between the kurtosis of the data sequences can reflect the similarity of the distribution between the data sequences. When the difference is large, it means that the data distribution of the two data sequences is very different, and the credibility of the large correlation between the two data sequences is low. is a preset hyperparameter, wherein the preset hyperparameter is set to prevent the denominator of the calculation formula of the comprehensive correlation from being zero. =0.01; is the absolute value symbol.

[0068] in, The smaller the The smaller the correlation between the voltage data and current data collected in the corresponding time period of the reference sequence of the missing data, the greater the degree of distortion of the data in the reference sequence of the missing data. At this time, the reason why the primary correction coefficient corresponding to the missing data is larger is that the credibility of the distorted data is higher, and the credibility of the real data is lower. Therefore, the primary correction coefficient corresponding to the missing data should be reduced to ensure the accuracy of the correction value of the prediction interval error factor of the missing data.

[0069] Step 3: Use the normalized value of the product of the correlation and the primary correction coefficient as the weight, and weight the prediction interval error factor of the missing data to obtain the correction value.

[0070] In an exemplary embodiment of the present invention, and The normalized value of the product of The final correction coefficient of the prediction interval error factor of missing data. Specifically, the final correction coefficient of the prediction interval error factor of any missing data satisfies the following relationship:

[0071] ;

[0072] In the formula, For the The final correction coefficient of the prediction interval error factor for missing data; For the The primary correction coefficient for the prediction interval error factor for missing data; For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; is the normalization function.

[0073] In another embodiment, a hyperbolic tangent function may be used for normalization. Of course, a suitable normalization method may be selected according to specific circumstances. This embodiment does not specifically limit the normalization method.

[0074] Furthermore, after determining the final correction coefficient of the prediction interval error factor of any missing data, the correction value of the prediction interval error factor of the missing data can be calculated. Specifically, the correction value of the prediction interval error factor of any missing data satisfies the following relationship:

[0075] ;

[0076] In the formula, For the Correction value of the prediction interval error factor for missing data; For the The final correction coefficient of the prediction interval error factor for missing data; For the It should be noted that the HTFE algorithm presets the same prediction interval error factor for all missing data in the voltage data sequence. =1.1.

[0077] By using the final correction coefficient of the prediction interval error factor of each missing data to correct the prediction interval error factor of the corresponding missing data, the prediction interval factor of each missing data can be adjusted in a targeted manner to avoid oversimplification caused by using a unified prediction interval factor. The predicted value at the corresponding moment of each missing data determined by the HTFE algorithm can better adapt to the volatility and abnormal characteristics of the data, and can effectively avoid the adaptive deviation caused by distorted data, thereby ensuring the accuracy of the interpolation result.

[0078] Next, the interpolation process of the missing data in the voltage data sequence is described in detail: for any missing data in the voltage data sequence, first, the correction value of the prediction interval error factor of the missing data is calculated, and then based on the correction value and the preset prediction error factor, the HTFE algorithm is used to predict the voltage data at the moment corresponding to the missing data, and the predicted value is used as the missing data, thereby realizing the interpolation of the missing data, and repeating the process to realize the interpolation of all missing data in the voltage data sequence to obtain a complete data sequence. It should be noted that, in this embodiment, when the HTFE algorithm is used to predict the voltage data at the moment corresponding to the missing data based on the correction value of the prediction error factor, the prediction error factor in the algorithm is 0.9. It should be noted that the process of determining the prediction value at each moment based on the prediction error factor and the prediction interval error factor using the HTFE algorithm is a prior art, and this embodiment will not be described in detail here.

[0079] S3: Perform anomaly detection on the complete data sequence and detect the welding status based on the anomaly detection results.

[0080] In an exemplary embodiment of the present invention, anomaly detection of a complete data sequence can be achieved by the following steps:

[0081] The LOF anomaly detection algorithm is used to perform anomaly detection on the complete data sequence.

[0082] In another embodiment, other anomaly detection algorithms may be used to perform anomaly detection on the complete data sequence, such as a box plot method, etc. This embodiment does not specifically limit the selected anomaly detection algorithm. It should be noted that the process of performing anomaly detection using the LOF anomaly detection algorithm is a prior art and is not described in detail in this embodiment.

[0083] In an exemplary embodiment of the present invention, the welding state can be determined by the following steps:

[0084] When any voltage data in the voltage data sequence is determined to be an abnormal value, the welding state of the electron beam at the time corresponding to the voltage data is judged to be an abnormal state; when any voltage data in the voltage data sequence is determined to be a normal value, the welding state of the electron beam at the time corresponding to the voltage data is judged to be a normal state.

[0085] Optionally, when it is determined that the welding state of the electron beam at any moment is abnormal, the technicians can be notified in time to perform maintenance, thereby ensuring the welding quality and improving production efficiency.

[0086] The present invention also provides an electron beam welding state detection system, the system includes a memory and a processor, and a computer program is stored in the memory, the computer program integrates the function of an electron beam welding state detection method, when the computer program is executed, the electron beam welding state can be accurately detected through an electron beam welding state detection method.

[0087] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0088] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for detecting electron beam welding status, characterized in that: include: Obtaining voltage data sequence during welding; The missing data in the voltage data sequence are interpolated using the HTFE algorithm. When interpolating each missing data, the correction value of the prediction interval error factor in the HTFE algorithm is calculated, and the corresponding missing data is replaced by the voltage data at the corresponding moment predicted based on the correction value to obtain a complete data sequence, so as to perform anomaly detection on the complete data sequence, and detect the welding state based on the anomaly detection result; The method for obtaining the correction value includes: taking a sequence consisting of several voltage data before any missing data as a reference sequence, calculating the primary correction coefficient of the prediction interval error factor of the missing data : ; , Respectively The missing data in the reference sequence The sample values ​​and predicted values ​​of the data; is the amount of data in the reference sequence; is the number of extreme points in the reference sequence; is the variance of the sampling intervals between adjacent extreme points in the reference sequence; is the absolute value symbol; Calculate the correlation of the voltage data and current data collected in the corresponding period of the reference sequence, take the normalized value of the product of the correlation and the primary correction coefficient as the weight, weight the prediction interval error factor of the missing data, and obtain the correction value; the correlation is a comprehensive correlation, and the comprehensive correlation satisfies the following relationship: ； In the formula, For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; For the The Pearson correlation coefficient of the voltage data and current data collected during the corresponding period of the reference sequence with missing data; , Respectively The kurtosis of the voltage data and the kurtosis of the current data collected during the corresponding period of the reference sequence with missing data; To preset hyper parameters; is the absolute value symbol.

2. The electron beam welding state detection method according to claim 1, characterized in that: The normalized value of the product of the comprehensive correlation and the primary correction coefficient is used as the final correction coefficient of the prediction interval error factor corresponding to the missing data. The final correction coefficient satisfies the following relationship: ； In the formula, For the The final correction coefficient of the prediction interval error factor for missing data; For the The primary correction coefficient for the prediction interval error factor for missing data; For the The comprehensive correlation of voltage data and current data collected during the corresponding period of the reference sequence with missing data; is the normalization function.

3. The electron beam welding state detection method according to claim 1, characterized in that: After acquiring the voltage data sequence during the welding process, the method further includes: The voltage data sequence is curve-fitted using the least square method to obtain a fitting curve of the voltage data sequence.

4. The electron beam welding state detection method according to claim 1, characterized in that: The method for obtaining the number of extreme value points in the reference sequence includes: The extreme value points in the reference sequence are identified by using a difference method, and the extreme value points in the reference sequence are counted in sequence to obtain the number of extreme value points in the reference sequence.

5. The electron beam welding state detection method according to claim 1, characterized in that: The anomaly detection on the complete data sequence includes: The LOF anomaly detection algorithm is used to perform anomaly detection on the complete data sequence.

6. The electron beam welding state detection method according to claim 1, characterized in that: The detecting of the welding state based on the abnormal detection result comprises: When it is determined that any voltage data in the voltage data sequence is an abnormal value, it is determined that the welding state of the electron beam at the time corresponding to the voltage data is an abnormal state; When it is determined that any voltage data in the voltage data sequence is a normal value, it is determined that the welding state of the electron beam at the time corresponding to the voltage data is a normal state.

7. An electron beam welding state detection system, characterized in that: The electron beam welding state detection system comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the electron beam welding state detection method as described in any one of claims 1-6.

Images