Welding data analysis method of welding platform based on big data

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a welding data analysis method for a welding platform based on big data, the method comprising: collecting parameter data of each welding data, obtaining the degree of isolation of the parameter data of each welding data, obtaining the local similarity of the parameter data of each welding data according to the local data of each welding data, obtaining the smoothing parameters of the parameter data of each welding data according to the local similarity and the degree of isolation of the parameter data of each welding data, and obtaining the predicted value of the parameter data of each welding data, obtaining the comprehensive abnormal performance of each welding data according to the deviation between the predicted value and the true value, and then judging whether there is a fault in the welding equipment at the sampling moment corresponding to each welding data. The present invention adapts the smoothing parameters of the parameter data to improve the accuracy of the predicted value, thereby improving the accuracy of the abnormality detection result.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more specifically, to a welding data analysis method for a welding platform based on big data. Background Art

[0002] In modern manufacturing, welding technology is a key process for connecting materials and building complex structures, especially in the fields of aerospace, automobile manufacturing, new energy, petrochemicals, marine engineering, etc., where welding quality directly affects the safety, reliability and service life of products. Laser welding, as a high-precision and high-efficiency welding method, has been widely used in various welding processes due to its advantages such as small heat-affected zone, high weld quality and small deformation. Among them, stable parameter data needs to be maintained during laser welding to ensure welding uniformity and form high-quality welds. Once the parameter data fluctuates or suddenly changes, it will have a significant impact on the welding quality. Therefore, the present invention proposes to monitor the welding data in real time and perform abnormal detection on each parameter data, and prompt adjustment in time when abnormalities are found, so as to improve the welding quality.

[0003] The patent application document currently published with the publication number CN112651774A is a method for optimizing the exponential smoothing model of a sales forecasting system, including: determining the vertices of a polyhedron, and constructing a polyhedron according to the vertices; finding the vertex with the largest function value, i.e., the worst point, and then discarding the worst point; calculating the center point of the remaining vertices, calculating the reflection point of each vertex, and determining whether the reflection point is within the feasible domain. If so, setting the current reflection coefficient value to a fixed value; if not, changing the reflection coefficient value and recalculating the reflection point until the reflection point is within the feasible domain of the polyhedron, and setting the current reflection coefficient value to a fixed value; comparing the objective function value of the reflection point with the objective function value of the worst point, and obtaining the optimal smoothing coefficient based on the comparison result.

[0004] It is known that in the actual laser welding process, the operating condition of the welding equipment is a key factor affecting the stability of the welding data. For example, factors such as wear and aging of internal components of the equipment and unstable power supply may cause sudden abnormalities in parameter data, thereby affecting the welding quality. In order to ensure the stability and consistency of the welding quality, it is necessary to perform abnormality detection on each parameter data. The exponential smoothing algorithm in the above method uses a fixed smoothing parameter, which is difficult to respond to the sudden changes in parameter data caused by the welding equipment in a timely manner, resulting in a large deviation between the predicted result and the actual value, reducing the accuracy of subsequent abnormality detection, resulting in false alarms or missed alarms, and causing adverse effects on the welding quality of the final product. Summary of the invention

[0005] In order to solve the problem that the traditional exponential smoothing algorithm adopts a fixed smoothing parameter, which is difficult to timely respond to the parameter data mutation caused by the welding equipment, resulting in a large deviation between the predicted result and the actual value, and reducing the accuracy of subsequent abnormality detection, the present invention proposes a welding data analysis method for a welding platform based on big data, which includes the following steps:

[0006] Collect each parameter data of each welding data; use the iterative self-organizing clustering algorithm to cluster each parameter data of all welding data, and obtain the cluster center of the cluster to which each parameter data of each welding data belongs in each clustering result; obtain the isolation degree of each parameter data of each welding data:

[0007] , Represents the degree of isolation of the ath parameter data of the ith welding data; represents the distance between the ath parameter data of the i-th welding data and the cluster center of the cluster to which it belongs in the k-th clustering result; K represents the number of clustering; norm() represents the normalization function;

[0008] Get the local similarity of each parameter data of each welding data:

[0009] , represents the local similarity of the ath parameter data of the i-th welding data; represents the mean value of all data in the local time series of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-1th welding data; norm() represents the normalization function; || represents the absolute value symbol; the ratio of the isolation degree of each parameter data of each welding data to the local similarity is recorded as the smoothing parameter of each parameter data of each welding data; based on the smoothing parameter, the predicted value of each parameter data of each welding data is obtained;

[0010] According to the deviation between the predicted value and the actual value of each parameter data of each welding data, the comprehensive abnormal performance of each welding data is obtained; based on the comprehensive abnormal performance, it is judged whether the welding equipment has a fault.

[0011] The innovation of the present invention lies in evaluating the isolation degree of each parameter data of each welding data through a dynamic clustering method, and obtaining the local similarity of each parameter data of each welding data in combination with the data change of the local time series, so as to obtain the smoothing parameters of each parameter data of each welding data according to the isolation degree and local similarity of each parameter data of each welding data, and predict the value of each parameter data of each welding data. Compared with fixed smoothing parameters, adaptive smoothing parameters can capture the dynamic changes of data caused by problems with the operating status of the equipment or other factors, significantly improve the accuracy of the prediction model, and make the obtained prediction value more accurate. Furthermore, when performing anomaly detection based on more accurate prediction values, false alarms and missed alarms are reduced, the accuracy of anomaly detection is improved, and the fault judgment of welding equipment is more accurate.

[0012] Preferably, the acquisition of the local time series of the parameter data includes:

[0013] The number of local data is preset to m, and m same parameter data before the sampling moment corresponding to the a-th parameter data of the i-th welding data are obtained as the local time series sequence of the a-th parameter data of the i-th welding data.

[0014] Preferably, the step of obtaining the predicted value of each parameter data of each welding data comprises:

[0015] According to the smoothing parameter of the ath parameter data of each welding data, the ath parameter data of all welding data are smoothed by using an exponential smoothing algorithm to obtain a predicted value of the ath parameter data of each welding data.

[0016] Preferably, the obtaining of comprehensive abnormal performance of each welding data includes:

[0017] Get historical data of each welding data;

[0018] ;

[0019] In the formula, Represents the comprehensive abnormal performance of the bth welding data; represents the value of the ath parameter data of the i-th welding data; represents the predicted value of the ath parameter data of the i-th welding data; Represents the number of dimensions; Represents the number of historical data of the i-th welding data; represents the value of the ath parameter data of the dth historical data of the ith welding data; Represents the predicted value of the ath parameter data of the dth historical data of the ith welding data; norm() represents the normalization function.

[0020] Preferably, the acquiring of historical data of each welding data comprises:

[0021] All welding data before the sampling time corresponding to each welding data is taken as the historical data of each welding data.

[0022] Preferably, judging whether the welding equipment has a fault based on the comprehensive abnormal performance includes:

[0023] The warning threshold T1 is preset. If the comprehensive abnormal performance of any welding data is greater than the warning threshold, the welding equipment may have a fault at the sampling time corresponding to the welding data, and a warning message is immediately sent to notify the operator to handle it in time.

[0024] Preferably, the acquisition of various parameter data of various welding data includes:

[0025] The preset sampling time is 2 seconds / time, and the data is collected for a total of 2 hours. During the welding process of the laser welding machine and the wire feeder, the wire feeding speed, laser power and welding speed are collected at the same time. The wire feeding speed, laser power and welding speed collected at each sampling time are recorded as each welding data, and each welding data contains multiple parameter data.

[0026] The present invention has the following beneficial effects: the purpose of the present invention is to evaluate the degree of isolation of each parameter data of each welding data through a dynamic clustering method, and obtain the local similarity of each parameter data of each welding data in combination with the data change of the local time series, so as to obtain the smoothing parameters of each parameter data of each welding data according to the degree of isolation and the local similarity of each parameter data of each welding data, and predict the value of each parameter data of each welding data. Compared with the fixed smoothing parameters, the adaptive smoothing parameters can capture the dynamic changes of data caused by problems with the operating status of the equipment or other factors, significantly improve the accuracy of the prediction model, and make the obtained prediction value more accurate. Furthermore, when performing anomaly detection based on more accurate prediction values, false alarms and missed alarms are reduced, the accuracy of anomaly detection is improved, and the fault judgment of the welding equipment is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. 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:

[0028] Figure 1 It is a flow chart of the steps of a welding data analysis method of a welding platform based on big data in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] 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.

[0030] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0031] See also Figure 1 , which shows a flowchart of a welding data analysis method for a welding platform based on big data provided by an embodiment of the present invention, the method comprising the following steps:

[0032] S001. Collect parameter data of each welding data.

[0033] In an embodiment of the present invention, the preset sampling time is 2 seconds / time. During the welding process of the laser welding machine and the wire feeder, the wire feeding speed, laser power and welding speed are collected simultaneously and recorded as each welding data. The data is collected for a total of 2 hours. Each welding data contains multiple parameter data of different types. The multiple parameter data include wire feeding speed, laser power and welding speed.

[0034] S002. Obtain the isolation degree of each parameter data of each welding data.

[0035] It should be noted that in the actual laser welding process, the operating condition of the welding equipment is a key factor affecting the stability of the welding data. For example, factors such as wear and aging of internal components of the equipment and unstable power supply may cause sudden abnormalities in the welding data, thereby affecting the welding quality. In order to ensure the stability and consistency of the welding quality, it is necessary to perform abnormality detection on the welding data. The traditional exponential smoothing algorithm uses fixed smoothing parameters, which makes it difficult to respond to the sudden changes in welding data caused by the welding equipment in a timely manner, resulting in a large deviation between the predicted results and the actual values, reducing the accuracy of subsequent abnormality detection, resulting in false alarms or missed alarms, and adversely affecting the welding quality of the final product.

[0036] It should be further explained that the parameter data of the welding data in the normal welding process should be stable to maintain the uniformity of the weld of the workpiece during the welding process, thereby ensuring the welding quality. However, if the internal components of the equipment are constantly worn or the power supply is unstable during the welding process, any parameter data may be abnormal. For example, there are abnormal fluctuations in the speed at which the welding wire is fed into the welding area, which will cause local accumulation of welding wire or uneven mixing, resulting in uneven welds. In order to accurately capture these anomalies, iterative clustering is performed on the arbitrary parameter data of all welding data to obtain the clustering results of each round. If the arbitrary parameter data of any welding data is far away from the cluster center of the cluster to which it belongs in each round of clustering results, it means that the parameter data of the welding data is more isolated in the distribution of the parameter data of all welding data.

[0037] In the embodiment of the present invention, a clustering number K is preset, and according to the clustering number K, an iterative self-organizing clustering algorithm is used to cluster each parameter data of all welding data, and the cluster center of the cluster cluster to which each parameter data of each welding data belongs in each clustering result is obtained;

[0038] Get the isolation degree of the ath parameter data of each welding data:

[0039] ;

[0040] In the formula, Represents the degree of isolation of the ath parameter data of the ith welding data; represents the distance between the ath parameter data of the i-th welding data and the cluster center of the cluster to which it belongs in the k-th clustering result; K represents the number of clustering; norm() represents the normalization function; Represents the sum of the distances between the ath parameter data of the ith welding data and the cluster center of the cluster to which it belongs in each clustering result. The larger the value, the more isolated the ath parameter data of the ith welding data is among all the ath parameter data. The fluctuation of the distance between the ath parameter data of the i-th welding data and the cluster center of the cluster to which it belongs in each clustering result is represented. The smaller the value, the more stable the ath parameter data of the i-th welding data is after multiple iterations. The larger the value of The smaller the value of , it means that the a-th parameter data of the i-th welding data is away from the cluster center to which it belongs after each clustering, and this distance is stable. Therefore, the degree of isolation of the a-th parameter data of the i-th welding data is greater.

[0041] S003. Obtain the local time series sequence of each parameter data of each welding data; obtain the local similarity of each parameter data of each welding data according to the local time series sequence of each parameter data of each welding data, obtain the smoothing parameters of each parameter data of each welding data according to the local similarity and isolation degree of each parameter data of each welding data, and obtain the predicted value of each parameter data of each welding data.

[0042] It should be noted that since the welding data of the welding process is continuous and usually stable in time series to ensure the uniformity of the weld, when the equipment condition changes, this stability will be broken, which is manifested as a sudden change in the data that significantly deviates from the normal range. If it is not adjusted in time, this sudden change may continue and intensify. The analysis of the degree of isolation only focuses on the abnormality of each parameter data of each welding data itself, but ignores the coherence and change trend of each parameter data of each welding data in time series, and thus cannot effectively capture such abnormal trend data. If any parameter data of a certain welding data has a significant deviation from the mean of the data in its local time series sequence, and its rate of change with the last data in the local time series sequence is also significant, it means that the parameter data of the welding data has significant differences in both value and trend. It is believed that the less similar the parameter data of the welding data is in the local time series, that is, the smaller the local similarity of the parameter data of the welding data is, the more likely the parameter data of the welding data is abnormal trend data.

[0043] In an embodiment of the present invention, m pieces of the same parameter data before the sampling moment corresponding to the a-th parameter data of the ith welding data are obtained as the local time series sequence of the a-th parameter data of the ith welding data; the number of local data m is preset to 10. In other embodiments, the implementer may preset the value of the number of local data m according to the specific implementation method.

[0044] Get the local similarity of the ath parameter data of each welding data:

[0045] ;

[0046] In the formula, represents the local similarity of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-th welding data; represents the mean value of all data in the local time series of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-1th welding data; norm() represents the normalization function; Represents the absolute difference between the ath parameter data of the i-th welding data and the mean of its local data. The larger the value, the more significant the deviation of the ath parameter data of the i-th welding data from the recent average value. It is believed that the ath parameter data of the i-th welding data is less similar in local time series. Represents the rate of change between the ath parameter data of the i-th welding data and the ath parameter data of the i-1-th welding data. The larger the value is, the more obvious the trend change is between the ath parameter data of the i-th welding data and the previous data, and the more dissimilar the ath parameter data of the i-th welding data is in the local time series.

[0047] It should be noted that the traditional exponential smoothing algorithm uses a fixed smoothing parameter to process all data, which is effective when the data changes relatively smoothly. When the data changes suddenly or the trend changes, if the smoothing parameter is too small, the model smoothes the current data by giving a larger weight to the historical data, resulting in the prediction results being more biased towards the historical normal data, failing to reflect the latest change trend of the current data in a timely manner, and reducing the accuracy of the prediction. Therefore, in order to make the prediction results more reliable, the size of the smoothing parameter is adjusted in real time according to the actual changes in the data, so that the model can better adapt to the change trend of the data. For example, after the data mutates and is in a range of continuous change, the smoothing parameter of the data is increased so that the change trend of the data in the local time series can be responded to more quickly when predicting the data, thereby improving the accuracy of the predicted value.

[0048] In the embodiment of the present invention, the smoothing parameter of the ath parameter data of each welding data is obtained:

[0049] ;

[0050] In the formula, represents the smoothing parameter of the ath parameter data of the i-th welding data; Represents the degree of isolation of the ath parameter data of the ith welding data; represents the local similarity of the ath parameter data of the ith welding data; norm() represents the normalization function; if the distribution of the ath parameter data of the ith welding data is more isolated and less similar in local time series, it means that the ath parameter data of the ith welding data is in the stage of mutation data or continuous change along the mutation trend caused by factors such as changes in equipment conditions during the welding process. At this time, a larger smoothing parameter should be given to quickly respond to the changing trend of local data, thereby improving the predicted value of such changing data.

[0051] According to the smoothing parameter of the ath parameter data of each welding data, the ath parameter data of all welding data are smoothed by using an exponential smoothing algorithm to obtain a predicted value of the ath parameter data of each welding data.

[0052] S004. Obtain the comprehensive abnormal performance of each welding data based on the deviation between the predicted value and the actual value of each parameter data of each welding data, and determine whether there is a fault in the welding equipment at the sampling moment corresponding to each welding data based on the comprehensive abnormal performance of each welding data.

[0053] It should be noted that if the deviation between the predicted value and the actual value of any parameter data of a certain welding data collected is significantly higher than the average of the deviation between the predicted value and the actual value of the parameter data of its historical data, this may mean that there is an abnormality in the parameter data of the welding data, and there may be various problems in the actual welding equipment during operation, such as electrode wear, unstable power supply, wire feeder equipment failure, etc. These problems will affect each parameter data in the welding process. In order to more accurately evaluate the operating status of the welding equipment, the abnormal degree of multiple parameter data of each welding data is considered, and the comprehensive abnormal performance of each welding data is obtained to comprehensively evaluate whether there is a fault in the equipment status at the sampling time corresponding to each welding data.

[0054] In the embodiment of the present invention, all welding data before the sampling moment corresponding to each welding data is taken as the historical data of each welding data.

[0055] Get comprehensive abnormal performance of each welding data:

[0056] ;

[0057] In the formula, Represents the comprehensive abnormal performance of the bth welding data; represents the value of the ath parameter data of the i-th welding data; represents the predicted value of the ath parameter data of the i-th welding data; Represents the number of dimensions; Represents the number of historical data of the i-th welding data; represents the value of the ath parameter data of the dth historical data of the ith welding data; represents the predicted value of the ath parameter data of the dth historical data of the ith welding data; Represents the abnormality of the ath parameter data of the i-th welding data. The larger the value, the greater the deviation between the actual value and the predicted value of the ath parameter data of the i-th welding data relative to the mean deviation of its historical data, and it is considered that the abnormality of the ath parameter data of the i-th welding data is higher. By summing the abnormalities of all parameter data of the i-th welding data, the comprehensive abnormal performance of the i-th welding data is obtained.

[0058] The preset warning threshold T1=0.8, if the comprehensive abnormal performance of any welding data is greater than the warning threshold, it is considered that the welding equipment at the sampling time corresponding to the welding data may have a fault resulting in abnormal data acquisition, and a warning message is sent to notify the operator to handle it in time. In the embodiment of the present invention, the implementer can preset the value of the warning threshold according to the specific implementation method.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A welding data analysis method for a welding platform based on big data, characterized in that: include: Collect various parameter data of various welding data; Use the iterative self-organizing clustering algorithm to cluster each parameter data of all welding data, and obtain the cluster center of the cluster to which each parameter data of each welding data belongs in each clustering result; obtain the isolation degree of each parameter data of each welding data: , Represents the degree of isolation of the ath parameter data of the ith welding data; represents the distance between the ath parameter data of the i-th welding data and the cluster center of the cluster to which it belongs in the k-th clustering result; K represents the number of clustering; norm() represents the normalization function; Get the local similarity of each parameter data of each welding data: , represents the local similarity of the ath parameter data of the i-th welding data; represents the mean value of all data in the local time series of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-th welding data; represents the value of the ath parameter data of the i-1th welding data; norm() represents the normalization function; || represents the absolute value symbol; the ratio of the isolation degree of each parameter data of each welding data to the local similarity is recorded as the smoothing parameter of each parameter data of each welding data; based on the smoothing parameter, the predicted value of each parameter data of each welding data is obtained; According to the deviation between the predicted value and the actual value of each parameter data of each welding data, the comprehensive abnormal performance of each welding data is obtained; Based on the comprehensive abnormal performance, it is determined whether the welding equipment has a fault.

2. The welding data analysis method of welding platform based on big data according to claim 1 is characterized in that: The acquisition of the local time series of the parameter data includes: The number of local data is preset to m, and m same parameter data before the sampling moment corresponding to the a-th parameter data of the i-th welding data are obtained as the local time series sequence of the a-th parameter data of the i-th welding data.

3. The welding data analysis method of welding platform based on big data according to claim 1, characterized in that: The step of obtaining the predicted value of each parameter data of each welding data comprises: According to the smoothing parameter of the ath parameter data of each welding data, the ath parameter data of all welding data are smoothed by using an exponential smoothing algorithm to obtain a predicted value of the ath parameter data of each welding data.

4. The welding data analysis method of welding platform based on big data according to claim 1, characterized in that: The comprehensive abnormal performance of each welding data is obtained, including: Get historical data of each welding data; ； In the formula, Represents the comprehensive abnormal performance of the bth welding data; represents the value of the ath parameter data of the i-th welding data; represents the predicted value of the ath parameter data of the i-th welding data; Represents the number of dimensions; Represents the number of historical data of the i-th welding data; represents the value of the ath parameter data of the dth historical data of the ith welding data; Represents the predicted value of the ath parameter data of the dth historical data of the ith welding data; norm() represents the normalization function.

5. The welding data analysis method of welding platform based on big data according to claim 4 is characterized in that: The obtaining of historical data of each welding data includes: All welding data before the sampling time corresponding to each welding data is taken as the historical data of each welding data.

6. The welding data analysis method of welding platform based on big data according to claim 1, characterized in that: The step of judging whether the welding equipment has a fault based on the comprehensive abnormal performance includes: The warning threshold T1 is preset. If the comprehensive abnormal performance of any welding data is greater than the warning threshold, the welding equipment may have a fault at the sampling time corresponding to the welding data, and a warning message is immediately sent to notify the operator to handle it in time.

7. The welding data analysis method of welding platform based on big data according to claim 1, characterized in that: The collecting of various parameter data of various welding data includes: The preset sampling time is 2 seconds / time, and the data is collected for a total of 2 hours. During the welding process of the laser welding machine and the wire feeder, the wire feeding speed, laser power and welding speed are collected at the same time. The wire feeding speed, laser power and welding speed collected at each sampling time are recorded as each welding data, and each welding data contains multiple parameter data.

Images