Luminescence sensing system for welding

• US 7,132,623 B2
• Filed: 03/27/2003
• Issued: 11/07/2006
• Est. Priority Date: 03/27/2002
• Status: Active Grant

• Current Assignee: Praxair Technology Incorporated (Linde Plc)
• Original Assignee: Praxair Technology Incorporated (Linde Plc)
• Inventors: Scotti, Americo, De Miranda, Helio Cordeiro, Ferraresi, Valtair Antonio
• Primary Examiner(s): Shaw, Clifford C.
• Application Number:
  US10/401,518
• Publication Number:
  US 20040034608A1
• Time in Patent Office: 1,321 Days
• Field of Search: 219/130.51, 219/130.5, 219/130.01
• US Class Current: 219/130.51
• CPC Class Codes: B23K 31/006 relating to using of neural networks
• CPC Class Codes: B23K 9/093 the frequency of the pulses produced being modulatable
• CPC Class Codes: B23K 9/0953 using computing means
• CPC Class Codes: B23K 9/0956 using sensing means, e.g. optical
• CPC Class Codes: B23K 9/1062 with computing means

Abstract

A method and apparatus can adaptively control a pulsed power arc welding process. A trainable system can recognize an empirical transfer mode from a signal emitted during an arc welding pulse and determine a pulsed power parameter set to produce a modified transfer mode in a subsequent pulse, by controlling a power source using the parameter set.

Claims

1. A method of adaptively controlling a pulsed power arc welding process, comprising:

sensing a signal emitted during an arc welding pulse;

employing a trainable system to;

recognize an empirical transfer mode from the signal;

determine a pulsed power parameter set to produce a modified transfer mode in a subsequent pulse; and

controlling a power source using the parameter set.

2. The method of claim 1, wherein the signal is a radiant flux signal.

3. The method of claim 2, whereby only a single metal droplet is transferred during each pulse for about 90% of the pulses.

4. The method of claim 3, wherein the parameter set comprises at least one value selected from the group consisting of pulse current, pulse voltage, base current, pulse period, frequency, and base period.

5. The method of claim 4, wherein the modified transfer mode is produced by controlling the pulse period.

6. The method of claim 5, wherein the empirical transfer mode comprises indefinite droplet transfer behavior.

7. The method of claim 5, wherein the trainable system is an artificial neural network.

8. A method of controlling a pulsed power arc-welding process, comprising:

sensing a radiant flux signal emitted during an arc welding pulse;

employing a neural network to;

recognize an empirical transfer mode from the signal;

determine a pulse period increment to produce a modified transfer mode in a subsequent pulse; and

controlling a power source using pulse period increment, whereby only a single metal droplet is transferred during each pulse for at least about 90% of the pulses.

9. A method of training a neural network for controlling a pulsed power arc welding process, comprising:

producing training data and validation data for a plurality of pulse examples, wherein each example includes;

a representation of a signal emitted during at least one arc welding pulse;

an empirical transfer mode for the pulse;

producing a set of control flags, wherein each flag includes;

a classification of an empirical transfer mode relative to a target transfer mode;

a control action for at least one value in a pulsed power parameter set; and

training a neural network using the training data, the validation data and the flags, whereby the network;

recognizes the empirical transfer mode 01 one or more arc welding pulses; and

applies the control action, thereby producing a modified transfer mode in a subsequent pulse.

10. The method of claim 9, wherein the signal is a radiant flux signal.

11. The method of claim l0, wherein the target transfer mode consists essentially of a single metal droplet transferred during each pulse for about 90% of the pulses.

12. The method of claim 11, wherein the parameter set comprises a value selected from the group consisting of pulse current, pulse voltage, base current, pulse period, frequency, and base period.

13. The method of claim 12, wherein the neural network produces the modified transfer mode by controlling the pulse period.

14. The method of claim 13, wherein the empirical transfer mode comprises indefinite droplet transfer behavior.

15. An apparatus for controlling a pulsed power source for arc welding, comprising:

a sensor;

a controller, comprising;

a signal acquisition module, whereby a signal from the sensor is acquired;

a trainable system, wherein the system is trained to;

recognize an empirical transfer mode from the signal;

control a parameter set for a pulsed power source, whereby a modified transfer mode is produced in a subsequent pulse; and

a control interface to control the power source using the parameter set.

16. The apparatus of claim 15, wherein the sensor comprises a photodiode.

17. The apparatus of claim 16, wherein the trainable system is a neural network.

18. The apparatus of claim 17, wherein the signal acquisition module comprises an amplifier and an analog to digital converter.

19. The apparatus of claim 18, wherein the parameter set comprises a value selected from the group consisting of pulse current, pulse voltage, base current, pulse period, frequency, and base period.

20. An apparatus for controlling a pulsed power source for arc welding, comprising a neural network trained to:

recognize an empirical transfer mode from a signal emitted during an arc welding process; and

determine a parameter set for an arc welding pulsed power source to produce a modified transfer mode.

21. A system for adaptively controlling a pulsed power arc welding process, comprising:

means to sense a signal emitted during an arc welding pulse;

a trainable system comprising;

means to recognize an empirical transfer mode from the signal;

means to determine a pulsed power parameter set to produce a modified transfer mode in a subsequent pulse; and

means to control a power source using the parameter set.

22. A system to control a pulsed power arc-welding process, comprising

means to sense a radiant flux signal emitted during an arc welding pulse;

a neural network comprising;

means to recognize an empirical transfer mode from the signal;

means to determine a pulsed power parameter set to produce a modified transfer mode in a subsequent pulse; and

means to control a power source using pulse period increment, whereby only a single metal droplet is transferred during each pulse for at least about 90% of the pulses.