Prediction method, device and system for consumable electrode pulse welding droplet transition and storage medium

Abstract

The invention discloses a consumable electrode pulse welding droplet transition prediction method, device and system and a storage medium, and relates to the technical field of welding process.The consumable electrode pulse welding droplet transition prediction method comprises the steps that pulse welding current waveform parameters are obtained; generating a current oscillogram according to the current waveform parameters, and exporting instantaneous current waveform data according to the current oscillogram; taking the instantaneous current waveform data as input, and performing prediction based on a voltage prediction model to obtain an instantaneous prediction voltage; and taking the instantaneous current waveform data and the instantaneous prediction voltage as input, and obtaining a molten drop transition moment based on the output of the molten drop transition model. According to the method, the droplet transfer time can be effectively predicted conveniently according to the current current waveform parameters, and the welding effect and the welding efficiency of the consumable electrode pulse welding process are improved.

Description

Technical Field

[0001] This invention relates to a method, apparatus, system, and storage medium for predicting droplet transfer in consumable electrode pulse welding, belonging to the field of welding process technology. Background Technology

[0002] With the development of welding technology, gas metal arc welding (GMAW) is increasingly being used in actual welding operations, and the technology is becoming more and more mature. Fully digital inverter welding power supplies are increasingly being used in industries such as engineering machinery, shipbuilding, and automobiles. These power supplies typically have a built-in set of expert data based on material, wire diameter, gas, and other conditions. However, this expert data is not always universally applicable to different industries, plate thicknesses, welding speeds, and shielding gases. Therefore, waveform parameters are usually adjusted to match the current welding scenario.

[0003] The consumable electrode pulse welding process is widely used due to its advantages such as minimal welding spatter, good weld formation, and high welding speed. The voltage and current waveforms of pulse welding are as follows: Figure 1 As shown, the upper part is the voltage waveform and the lower part is the current waveform. Before welding, it is usually necessary to pre-set the current and voltage, which are periodic average current and voltage.

[0004] After setting the current and voltage, the welding effect will vary greatly when welding different plate thicknesses, using different proportions of shielding gas, or welding speeds. For example, when welding carbon steel using shielding gases of 92%Ar+8%CO2 and 82%Ar+18%CO2, the welding arc length and the amount of spatter produced will be different. In this case, it is necessary to fine-tune the waveform parameters, which makes the welding operation too cumbersome. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, device, system and storage medium for predicting droplet transfer in consumable electrode pulse welding, which facilitates effective prediction of droplet transfer time based on the current current waveform parameters, thereby improving the welding effect and welding efficiency of the consumable electrode pulse welding process.

[0006] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides a method for predicting droplet transfer in consumable electrode pulse welding, comprising: S1. Obtain the current waveform parameters for pulse welding; S2. Generate a current waveform diagram based on the current waveform parameters, and export instantaneous current waveform data based on the current waveform diagram; S3. Using instantaneous current waveform data as input, the instantaneous predicted voltage is obtained based on the voltage prediction model; S4. Using instantaneous current waveform data and instantaneous predicted voltage as input, the droplet transition time is obtained based on the droplet transition model. S5. Determine the droplet transition time. If the droplet transition time is within the set time period, the droplet transition is deemed appropriate. If the droplet transition time is not within the set time period, adjust the current waveform parameters of the pulse welding and repeat S2~S5 until the droplet transition is deemed appropriate, thus obtaining the predicted droplet transition time.

[0007] Furthermore, the current waveform parameters include time parameters and current parameters. The current parameters include peak current, base current, second base current, rising slope, and falling slope. The time parameters include rising time period, peak maintenance time period, falling time period, and base maintenance time period.

[0008] Furthermore, the voltage prediction model is one of the following: a feedforward neural network model, a gated recurrent unit model, a combination model of convolutional neural network and long short-term memory neural network, and a train communication network model.

[0009] Furthermore, the method also includes pre-training the voltage prediction model, wherein the pre-training method includes: Acquire current and voltage training set data, which includes multiple data pairs consisting of actual current waveform data and their corresponding actual voltage data; The actual current waveform data is used as input to train the voltage prediction model until the deviation between the predicted voltage data output by the voltage prediction model and the actual voltage data is less than a first set threshold, thus obtaining the pre-trained voltage prediction model.

[0010] Furthermore, the droplet transition model is one of the following: long short-term memory neural network model, Transformer, and sequence-to-sequence model.

[0011] Furthermore, it also includes pre-training the molten droplet transition model, the pre-training method comprising: Acquire droplet transition training set data, which includes multiple data pairs consisting of actual current waveform data, actual voltage data, and actual droplet transition time; Using actual current waveform data and actual voltage data as input, the droplet transition model is trained until the deviation between the droplet transition time output by the droplet transition model and the actual droplet transition time is less than a second set threshold, thus obtaining a pre-trained droplet transition model.

[0012] Furthermore, it also includes judging the droplet transition time. If the droplet transition time is within the set time period, it is judged that the droplet transition meets the welding conditions. If the droplet transition time is not within the set time period, the current waveform parameters of pulse welding are adjusted. The set time period is the last 2 / 3 of the falling time period and the first 1 / 3 of the base value maintenance time period.

[0013] Secondly, the present invention also provides a predictive device for droplet transfer in consumable electrode pulse welding, comprising: The current waveform parameter acquisition module is configured to acquire the current waveform parameters for pulse welding. The instantaneous current waveform generation module is configured to generate a current waveform diagram based on the current waveform parameters, and to export instantaneous current waveform data based on the current waveform diagram. The instantaneous voltage prediction module is configured to take instantaneous current waveform data as input and predict the instantaneous voltage based on the voltage prediction model. The droplet transition time prediction module is configured to take instantaneous current waveform data and instantaneous predicted voltage as inputs and output the droplet transition time based on the droplet transition model.

[0014] Thirdly, the present invention also provides a computer-readable storage medium, wherein when the computer program is executed by a processor, it implements the method for predicting droplet transfer in molten electrode pulse welding as described in any of the first aspects.

[0015] Fourthly, the present invention also provides a computer system, comprising: Memory, used to store computer programs; A processor for executing the computer program to implement the method for predicting droplet transfer in molten electrode pulse welding as described in any of the first aspects.

[0016] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: This invention first predicts the voltage based on the current, and then predicts the droplet transition time based on the input current waveform parameters and the predicted voltage. This facilitates the effective prediction of the droplet transition time based on the current current waveform parameters, thereby improving the welding effect and welding efficiency of the consumable electrode pulse welding process. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the voltage and current waveforms for consumable electrode pulse welding. Figure 2 This is a flowchart illustrating a method for predicting droplet transfer in consumable electrode pulse welding according to one embodiment of the present invention. Figure 3 This is a schematic diagram of the current waveform of a method for predicting droplet transfer in consumable electrode pulse welding according to an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the predicted voltage with the original voltage in a method for predicting droplet transfer in consumable electrode pulse welding according to an embodiment of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention. Example 1

[0019] The pulse welding position constant current control method involves automatically generating the voltage based on the loop resistance during welding by inputting the waveform parameters of the current. Under normal circumstances, the loop resistance remains relatively constant, and the current magnitude determines the welding voltage. The energy generated by the welding current and voltage in one cycle determines the size of the molten droplet and the timing of its transition. Therefore, this invention provides a method for predicting the droplet transition in consumable electrode pulse welding, such as... Figure 2 As shown, it includes the following steps: S1. Obtain the current waveform parameters for pulse welding, such as... Figure 3 As shown, the current waveform adoption count consists of two parts: time parameters and current parameters. The current parameters include peak current, base current, second base current, rising slope, and falling slope. The time parameters include rising time period, peak maintenance time period, falling time period, and base maintenance time period.

[0020] S2. After inputting the current waveform parameters, generate the corresponding current waveform diagram according to the predetermined parameters, and export the instantaneous current waveform data from the current waveform diagram.

[0021] S3. Using the instantaneous current waveform parameters as input, the instantaneous predicted voltage is obtained based on the output of the voltage prediction model. The voltage prediction model can be any one of the following: feedforward neural network model, gated recurrent unit model, convolutional neural network and long short-term memory neural network combination model, and train communication network model, but is not limited to these. Any other network model that can achieve prediction should be acceptable.

[0022] In this embodiment, the voltage prediction model adopts a feedforward neural network model, and it is pre-trained first. The pre-training method includes: Acquire current and voltage training set data, which includes multiple data pairs consisting of actual current waveform data and their corresponding actual voltage data.

[0023] The actual current waveform data is used as input to train the voltage prediction model until the deviation between the predicted voltage data output by the voltage prediction model and the actual voltage data is less than a first set threshold, thus obtaining the pre-trained voltage prediction model. Figure 4 As shown, Figure 4The graph shows a comparison between the predicted voltage and the actual voltage (i.e., the original voltage in the figure). It can be observed from the graph that the deviation between the predicted voltage and the actual voltage is very small, which meets the accuracy requirements.

[0024] S4. Using the instantaneous current waveform data and instantaneous predicted voltage as input, the droplet transition time is obtained based on the droplet transition model. The droplet transition model can be any one of the following: long short-term memory neural network model, Transformer, and sequence-to-sequence model, but is not limited to these. Other network models that can achieve prediction should also be acceptable.

[0025] In this embodiment, the droplet transition model adopts a long short-term memory neural network model, which needs to be pre-trained. The pre-training method includes: Acquire droplet transition training set data, which includes multiple data pairs consisting of actual current waveform data, actual voltage data, and actual droplet transition time.

[0026] The actual current waveform data and actual voltage data are used as inputs to train the droplet transfer model until the deviation between the droplet transfer time output by the model and the actual droplet transfer time is less than a second preset threshold, thus obtaining a pre-trained droplet transfer model. The first and second preset thresholds are selected based on the actual welding conditions.

[0027] After obtaining the predicted droplet transition time, the process also includes judging the droplet transition time. If the droplet transition time is within the set time period, the droplet transition is judged to be appropriate. If the droplet transition time is not within the set time period, the current waveform parameters of pulse welding are adjusted, and the above prediction method is repeated according to the adjusted current waveform parameters until the droplet transition time meets the welding conditions.

[0028] In this embodiment, the time period is set to the latter 2 / 3 of the decreasing time period and the former 1 / 3 of the base value maintenance time period. Example 2

[0029] This embodiment provides a device for predicting droplet transfer in consumable electrode pulse welding, comprising: The current waveform parameter acquisition module is configured to acquire the current waveform parameters for pulse welding.

[0030] The instantaneous current waveform generation module is configured to generate a current waveform diagram based on the current waveform parameters, and to export instantaneous current waveform data based on the current waveform diagram.

[0031] The instantaneous voltage prediction module is configured to take instantaneous current waveform data as input and predict the instantaneous voltage based on the voltage prediction model.

[0032] The droplet transition time prediction module is configured to take instantaneous current waveform data and instantaneous predicted voltage as inputs and output the droplet transition time based on the droplet transition model. Example 3

[0033] This embodiment provides a computer-readable storage medium, which, when executed by a processor, implements the method for predicting droplet transfer in molten electrode pulse welding as described in Embodiment 1. Example 4

[0034] This embodiment provides a computer system, including: Memory, used to store computer programs; A processor is used to execute the computer program to implement the method for predicting droplet transfer in molten electrode pulse welding as described in Example 1.

[0035] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting droplet transfer in consumable electrode pulse welding, characterized in that, include: S1. Obtain the current waveform parameters for pulse welding; S2. Generate a current waveform diagram based on the current waveform parameters, and export instantaneous current waveform data based on the current waveform diagram; S3. Using instantaneous current waveform data as input, the instantaneous predicted voltage is obtained based on the voltage prediction model; S4. Using instantaneous current waveform data and instantaneous predicted voltage as input, the droplet transition time is obtained by outputting the droplet transition model.

2. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 1, characterized in that, The current waveform parameters include time parameters and current parameters. The current parameters include peak current, base current, second base current, rising slope and falling slope. The time parameters include rising time period, peak maintenance time period, falling time period and base maintenance time period.

3. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 1, characterized in that, The voltage prediction model is one of the following: feedforward neural network model, gated recurrent unit model, convolutional neural network and long short-term memory neural network combined model, and train communication network model.

4. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 3, characterized in that, It also includes pre-training the voltage prediction model, the pre-training method comprising: Acquire current and voltage training set data, which includes multiple data pairs consisting of actual current waveform data and their corresponding actual voltage data; The actual current waveform data is used as input to train the voltage prediction model until the deviation between the predicted voltage data output by the voltage prediction model and the actual voltage data is less than a first set threshold, thus obtaining the pre-trained voltage prediction model.

5. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 1, characterized in that, The droplet transition model is one of the following: long short-term memory neural network model, Transformer, and sequence-to-sequence model.

6. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 5, characterized in that, It also includes pre-training the molten droplet transition model, the pre-training method comprising: Acquire droplet transition training set data, which includes multiple data pairs consisting of actual current waveform data, actual voltage data, and actual droplet transition time; Using actual current waveform data and actual voltage data as input, the droplet transition model is trained until the deviation between the droplet transition time output by the droplet transition model and the actual droplet transition time is less than a second set threshold, thus obtaining a pre-trained droplet transition model.

7. The method for predicting droplet transfer in consumable electrode pulse welding according to claim 1, characterized in that, It also includes judging the droplet transition time. If the droplet transition time is within the set time period, it is judged that the droplet transition meets the welding conditions. If the droplet transition time is not within the set time period, the current waveform parameters of pulse welding are adjusted. The set time period is the last 2 / 3 of the falling time period and the first 1 / 3 of the base value maintenance time period.

8. A predictive device for droplet transfer in consumable electrode pulse welding, characterized in that, include: The current waveform parameter acquisition module is configured to acquire the current waveform parameters for pulse welding. The instantaneous current waveform generation module is configured to generate a current waveform diagram based on the current waveform parameters, and to export instantaneous current waveform data based on the current waveform diagram. The instantaneous voltage prediction module is configured to take instantaneous current waveform data as input and predict the instantaneous voltage based on the voltage prediction model. The droplet transition time prediction module is configured to take instantaneous current waveform data and instantaneous predicted voltage as inputs and output the droplet transition time based on the droplet transition model.

9. A computer-readable storage medium, characterized in that, When the computer program is executed by the processor, it implements the method for predicting droplet transfer in molten electrode pulse welding as described in any one of claims 1 to 7.

10. A computer system, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the method for predicting droplet transfer in molten electrode pulse welding as described in any one of claims 1 to 7.

Images