A GTAW droplet transfer uniformity real-time monitoring method based on arc sensing

Abstract

The present application relates to a kind of GTAW drop transfer uniformity real-time monitoring method based on arc sensing.The uniformity of drop transfer can reflect the uniformity of weld appearance size in welding process, which can be used for real-time monitoring of the uniformity of weld appearance size, and a kind of GTAW drop transfer uniformity real-time monitoring method based on arc sensing is disclosed for this characteristic.Wavelet transform-based arc signal filtering processing method is used to filter the original arc voltage signal;OTSU-based drop growth process feature signal extraction method is used to extract the drop growth process feature signal in arc signal;Drop transfer feature signal extraction method based on DBSCAN clustering is used to extract drop transfer feature signal;Drop transfer uniformity calculation method is used to obtain drop transfer uniformity.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of real-time monitoring of molten droplet transition in welding process, and is a real-time monitoring method for GTAW molten droplet transition uniformity based on arc sensing. BACKGROUND

[0002] Gas tungsten arc welding (GTAW) is widely used in important industrial fields such as aerospace, marine equipment, and nuclear power construction, and is mainly implemented by manual and semi-automatic welding. In the GTAW process, the molten droplet transition frequency as an important parameter plays a decisive role in the welding quality. The weld uniformity is affected by the molten droplet transition uniformity, and real-time monitoring of the molten droplet transition uniformity during the welding process can be used to control the weld forming uniformity. However, there is currently no method for real-time monitoring of the molten droplet transition uniformity. In view of the problem that the molten droplet transition uniformity in the GTAW welding process is difficult to monitor in real time, the present application discloses a real-time monitoring method for GTAW molten droplet transition uniformity based on arc sensing. Based on arc sensing, the GTAW molten droplet transition uniformity is monitored in real time by extracting and calculating the information of the welding arc, the welding process is monitored, and the welding quality is improved. SUMMARY

[0003] A real-time monitoring method for GTAW molten droplet transition uniformity based on arc sensing is used to monitor the molten droplet transition uniformity in the welding process, characterized in that the real-time monitoring method for GTAW molten droplet transition uniformity based on arc sensing is realized by a real-time monitoring system for molten droplet transition uniformity based on arc sensing; as shown in Figure 1 , the real-time monitoring system for molten droplet transition uniformity based on arc sensing is composed of a data acquisition system and a data processing system, the data acquisition system is composed of an arc voltage sensor and a data acquisition card, and is used to periodically acquire the arc signal in real time during the welding process , Figure 2 as shown in the figure, the arc signal is the voltage signal between the tungsten electrode and the workpiece; the data processing system is used for real-time processing of the arc signal, and realizes real-time monitoring of the GTAW molten droplet transition uniformity.

[0004] The flow chart of a real-time monitoring method for GTAW molten droplet transition uniformity based on arc sensing is shown in Figure 3 , the arc signal is processed by using an OTSU-based molten droplet growth process feature signal extraction method; all arc signal values in are arranged in ascending order, and the sorted values are obtained , and the frequency and frequency of all arc signal values are counted Assuming a threshold ( =1,2,3,……, ),get Center front The sum of the frequencies of each arc signal value and the One to the first The sum of the frequency values ​​of each arc signal value ,forward The average frequency of each arc signal value and the One to the first The average frequency of each arc signal value To determine the inter-class variance ; between-class variance Let the objective function be the assumed threshold. Traversal ,Compare Take different The magnitude of the inter-class variance, where the arc signal value that maximizes the inter-class variance. The corresponding arc signal threshold The aforementioned arc signal threshold This is used to separate the droplet growth process characteristic signal and the droplet-molten pool contact process signal in the arc signal. The arc signal above the arc signal threshold is the droplet growth process characteristic signal, and the arc signal below the arc signal threshold is the droplet-molten pool contact process signal. hour, As a signal of the droplet growth process, when hour, This is a signal indicating the contact process between the molten droplet and the molten pool. Figure 4 This image shows the segmentation results of the droplet transition characteristic signal and the droplet-molten pool contact process signal after determining the arc signal threshold. The droplet growth process characteristic signal refers to the arc signal formed when the welding wire melts and forms a droplet during GTAW welding, but the droplet has not yet contacted the molten pool; this signal remains at a high numerical level. The droplet-molten pool contact process signal refers to the arc signal formed as the droplet gradually approaches and contacts the molten pool, finally leaving the welding wire and fully entering the molten pool. The arc signal value corresponding to this stage exhibits a characteristic of first decreasing and then increasing.

[0005] The droplet transition feature signal is extracted using a DBSCAN clustering-based method. The droplet transition feature signal refers to the droplet-molten pool contact process signal corresponding to a single droplet transition process. The specific steps of the DBSCAN clustering-based droplet transition feature signal extraction method are as follows:

[0006] Step 1: Signals of the contact process between the molten droplet and the molten pool Obtained by formula (1)

[0007] (1)

[0008] Wherein, is the arc signal data of a sampling period, is the signal feature data of the droplet growth process, leaving all the droplet and molten pool contact process signals in a sampling period , as shown in Figure 5 is the droplet and molten pool contact process signal diagram.

[0009] Second step: extract the droplet transition feature signal

[0010] In a sampling period of arc signal, there are times of droplet transition, that is, the arc voltage signal data in the arc voltage sampling signal contains times of droplet transition arc voltage signal data. The droplet and molten pool contact process signal has certain characteristics in , as shown in Figure 5 , the characteristics are that all the droplet and molten pool contact process signal points of a droplet transition in are closely gathered together, and there is a large distance between the adjacent two droplet and molten pool contact process signal points. Based on the characteristics, the DBSCAN clustering method based on density clustering is used to process the droplet and molten pool contact process signal to extract the droplet transition feature signal. The DBSCAN clustering algorithm divides the arc signal into clusters and a noise point set, wherein the number of droplet transitions , and the droplet and molten pool contact process arc signal feature is as shown in formula (2):

[0011] (2)( is the cluster)

[0012] The droplet transition uniformity is calculated by using the droplet transition uniformity calculation method, and the specific steps of the droplet transition uniformity calculation method are as follows:

[0013] Suppose that there are droplet transition feature signals in a sampling period of arc signal , the data of the th droplet transition feature signal is , the corresponding data number is , and the time point corresponding to the th data point in the th droplet transition feature signal is .

[0014] (1) Calculate one arc signal sampling period using equation (3). Inner Centroid of the data point corresponding to the molten droplet transition characteristic signal:

[0015] (3)

[0016] (2) Calculate one arc signal sampling period using equation (4). The distance between the centroids of two adjacent droplet transition characteristic signals:

[0017] (4)

[0018] (3) Calculate one arc signal sampling period using equation (5). The average distance between the centroids of all two adjacent droplet transition characteristic signals within the droplet:

[0019] (5)

[0020] (4) Calculate one arc signal sampling period using equation (6). Variance of the centroid distance between two adjacent droplet transition characteristic signals :

[0021] (6)

[0022] like The larger the value, the less uniform the droplet transition, i.e., the lower the droplet transition uniformity; if The smaller the value, the closer it is to zero, the more uniform the droplet transition, that is, the higher the uniformity of the droplet transition.

[0023] The extracted droplet transition characteristic signal results are as follows: Figure 6 As shown.

[0024] Beneficial effects of the invention:

[0025] This invention relates to a real-time monitoring method for droplet transition uniformity in GTAW (Gross-Tapered Flow Arithmetic) based on arc sensing. It utilizes an OTSU-based droplet growth process feature signal extraction method to distinguish between droplet growth process feature signals and droplet-molten pool contact process signals; it extracts droplet transition feature signals using a DBSCAN clustering-based method; and it calculates droplet transition uniformity in real time using a droplet transition uniformity calculation method. This invention overcomes the difficulty of real-time monitoring of droplet transition uniformity during GTAW. Attached Figure Description

[0026] Figure 1 This is a system diagram of a real-time monitoring method for the uniformity of GTAW droplet transition based on arc sensing.

[0027] Figure 2 Arc signal graph acquired by the data acquisition system.

[0028] Figure 3 OTSU-based arc signal threshold selection flowchart for droplet growth process feature signal extraction method.

[0029] In the figure: - the value corresponding to each arc signal point, , - the arc signal values that have appeared, - the arc signal that has appeared, - the arc signal that has appeared, , - to , and - to , and - to , and - to , and - the inter-class variance function, - the arc signal threshold value to be selected, - the inter-class variance value when the arc signal threshold value to be selected is , - the maximum inter-class variance, - the arc signal threshold value.

[0030] Figure 4 Segmentation result graph of the droplet transfer feature signal and the droplet-pool contact process signal.

[0031] Figure 5 Droplet-pool contact process signal graph.

[0032] Figure 6 DBSCAN clustering-based droplet transfer feature signal extraction method result graph.

[0033] Figure 7 GTAW droplet transfer uniformity real-time monitoring method based on arc sensing flowchart. DETAILED DESCRIPTION

[0034] As shown in Figure 7 , a GTAW droplet transfer uniformity real-time monitoring method based on arc sensing flowchart, the specific steps are as follows:

[0035] The first step: the data acquisition system acquires the arc signal in the welding process and transmits to the data processing system;

[0036] The second step: the data processing system processes the arc signal data acquired in the first step in real time: the arc voltage signal is processed by using the OTSU-based droplet growth process characteristic signal extraction method, the arc signal threshold distinguishing the droplet growth process signal and the droplet-melt pool contact process signal is obtained, and the two are distinguished, the droplet growth process characteristic signal and the droplet-melt pool contact process signal are extracted respectively; the droplet transfer characteristic signal extraction method based on DBSCAN clustering is used to process the droplet-melt pool contact process signal, the number of droplet transfers represented by the droplet-melt pool contact process signal is obtained, and the droplet-melt pool contact process signals of the same droplet transfer are clustered into a droplet transfer characteristic signal;

[0037] The third step: the droplet transfer uniformity is calculated by using the droplet transfer uniformity calculation method, and the calculation result is output.

Claims

1. A GTAW droplet transfer uniformity real-time monitoring method based on arc sensing, for real-time monitoring of droplet transfer uniformity in a welding process, characterized in that: The aforementioned real-time monitoring method for GTAW droplet transfer uniformity based on arc sensing is implemented through a real-time monitoring system for droplet transfer uniformity based on arc sensing. This system comprises a data acquisition system and a data processing system. The data acquisition system consists of an arc voltage sensor and a data acquisition card, used to periodically acquire the arc signal V = [x1, x2, x3, ..., x...] in real time during the welding process. n The arc signal is the voltage signal between the tungsten electrode and the workpiece; the data processing system is used for real-time processing of the arc signal; the real-time monitoring method for GTAW droplet transition uniformity based on arc sensing uses an OTSU-based droplet growth process feature signal extraction method to process the arc signal V, obtain an arc signal threshold that distinguishes between the droplet growth process signal and the droplet-molten pool contact process signal, and then distinguishes between the two, extracting the droplet growth process feature signal and the droplet-molten pool contact process signal respectively; the droplet growth process feature signal is the arc formed when the welding wire melts to form a droplet during GTAW welding, and the droplet does not contact the molten pool. The signal remains at a high numerical level; the signal representing the contact process between the molten droplet and the molten pool is the arc signal formed during the process of the molten droplet gradually approaching and contacting the molten pool, and finally leaving the welding wire and completely entering the molten pool. The arc signal corresponding to this stage exhibits a characteristic of first decreasing and then increasing; the molten droplet transition feature signal is extracted using a DBSCAN clustering-based molten droplet transition feature signal extraction method; the molten droplet transition feature signal is the molten droplet contact process signal corresponding to one molten droplet transition process; the molten droplet transition uniformity is calculated using the molten droplet transition uniformity calculation method, by calculating the variance of the centroid distance between two adjacent molten droplet transition feature signals over one arc signal sampling period T. like The larger the value, the less uniform the droplet transition, i.e., the lower the droplet transition uniformity; if The smaller the value, the closer it is to zero, the more uniform the droplet transition, that is, the higher the droplet transition uniformity, thus enabling real-time monitoring of the GTAW droplet transition uniformity.

2. The GTAW droplet transfer uniformity real-time monitoring method based on arc sensing of claim 1, wherein: The OTSU-based droplet growth process characteristic signal extraction method takes the inter-class variance as a target function, and substitutes different arc signal values x i The arc signal value that maximizes the inter-class variance is the corresponding arc signal threshold value; the arc signal threshold value is used to separate the droplet growth process characteristic signal and the droplet-molten pool contact process signal in the arc signal, the arc signal higher than the arc signal threshold value is the droplet growth process characteristic signal, and the arc signal lower than the arc signal threshold value is the droplet-molten pool contact process signal.

3. The GTAW droplet transfer uniformity real-time monitoring method based on arc sensing of claim 1, wherein: The specific steps of the droplet transfer feature signal extraction method based on DBSCAN clustering are as follows: First step: the signal D of the contact process of the droplet and the molten pool is obtained through formula (1) Wherein, V is the arc signal data of a sampling period, V g is the signal characteristic data of the droplet growth process, and D is the signal of the contact process between all the droplets and the molten pool in a sampling period. Second step: extracting the droplet transfer feature signal N times of droplet transfers are carried out in one arc signal sampling period, that is, the arc signal data of N times of droplet transfers are contained in the arc sampling signal; the droplet-pool contact process signal has certain characteristics in D, the characteristics being that all the droplet-pool contact process signal points of one droplet transfer are closely gathered together in D, and there is a large distance between the signal points of two adjacent droplet transfers; based on the characteristics, the DBSCAN clustering method based on density clustering is used to process the droplet-pool contact process signal to extract the droplet transfer characteristic signal; the DBSCAN clustering algorithm divides the arc signal into c clusters and a noise point set, wherein the number of droplet transfers N = c, and the droplet-pool contact process arc signal characteristics D i As shown in formula (2): D i = C i , 1≤i≤c (2)(C i is the ith cluster).

4. The method of claim 1, wherein the method is characterized by: The specific steps of the droplet transfer uniformity calculation method are as follows: There are C arc signal sampling periods T in total, and the data of the i-th droplet transfer characteristic signal is V c i The corresponding data number is The time point corresponding to the k-th data point in the i-th droplet transfer characteristic signal is (1) The centroid of the data point corresponding to the i-th droplet transfer feature signal in an arc signal sampling period T is calculated by formula (3): (2) The distance between the centroids of two adjacent droplet transfer feature signals in an arc signal sampling period T is calculated by formula (4): (3) The average value of the distances between the centroids of all adjacent droplet transfer feature signals in an arc signal sampling period T is calculated by formula (5): (4) calculating a variance of the distance between the centers of two adjacent droplet transition characteristic signals in an arc signal sampling period T using formula (6) like The larger the value, the less uniform the droplet transition, i.e., the lower the droplet transition uniformity; if The smaller the value, the closer it is to zero, the more uniform the droplet transition, that is, the higher the uniformity of the droplet transition.

Images