Machine learning device having function of adjusting welding positions of core in wire electric discharge machine

• US 10,589,368 B2
• Filed: 07/11/2019
• Issued: 03/17/2020
• Est. Priority Date: 11/30/2015
• Status: Active Grant

• Current Assignee: Fanuc Corporation
• Original Assignee: Fanuc Corporation
• Inventors: Shirai, Kenichiro
• Primary Examiner(s): Tran, Dzung
• Application Number:
  US16/508,328
• Publication Number:
  US 20190329341A1
• Time in Patent Office: 250 Days
• Field of Search: 700162
• CPC Class Codes: B23H 5/00 Combined machining
• CPC Class Codes: B23H 7/02 Wire-cutting
• CPC Class Codes: B23H 7/20 for programme-control, e.g. adaptive programme-control systems in general G05B19/00
• CPC Class Codes: B23H 9/00 Machining specially adapted for treating particular metal objects or for obtaining special effects or results on metal objects heat treatment by cathodic discharge C21D1/38
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 19/25 for continuous-path control
• CPC Class Codes: G05B 19/402 characterised by control arrangements for positioning, e.g. centring a tool relative to a hole in the workpiece, additional detection means to correct position G05B19/19 takes precedence
• CPC Class Codes: G05B 2219/45221 Edm, electrical discharge machining, electroerosion, ecm, chemical
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A machine learning device, performing machine learning for adjusting a position and a length of a welding part when a core is welded to a workpiece in a wire electric discharge machine, acquires the position and the length of the welding part as state data; sets reward conditions; calculates a reward based on the state data and the reward conditions; performs the machine learning of the adjustment; determines and outputs an adjustment target and its adjustment amounts based on the state data and a result of the machine learning; performs the machine learning of the adjustment based on the output adjustment action, the state data acquired based on the recalculated position and the recalculated length of the welding part, and the reward based on the state data; and outputs an optimum position of the welding part, the reward conditions being set as a positive or negative reward.

Claims

1. A machine learning device that performs machine learning of an adjustment of a position and a length of a welding part formed on a machining path for machining to weld a core to a workpiece when the core is cut out from the workpiece by a wire electric discharge machine that performs the machining to cut out the core from the workpiece based on machining preconditions including a program, the machine learning device comprising:

a processor configured to;

acquire, as state data on the welding part, the position and the length of the welding part and an evaluation value to evaluate the position and the length of the welding part;

set reward conditions;

calculate a reward based on the state data and the reward conditions;

perform the machine learning of an adjustment of the position and the length of the welding part; and

determine and output (i) an adjustment target including at least one of the position and the length of the welding part and (ii) adjustment amounts of the adjustment target as an adjustment action, based on the state data and a result of the machine learning,

perform the machine learning of the adjustment of the position and the length of the welding part based on

(a) the output adjustment action,

(b) the state data acquired based on the recalculated position and the recalculated length of the welding part, and

(c) the reward calculated based on the state data acquired based on the recalculated position and the recalculated length of the welding part and

upon completion of the machine learning, output an optimum position of the welding part,

wherein

the reward conditions are set

such that the processor is configured to calculate a positive reward

(a) when a number of the welding parts is small or

(b) when a position for supporting the core is well balanced and such that the processor is configured to calculate a negative reward

(c) when the number of the welding parts is large, or

(d) when the length of the welding part is shorter than a previously-set welding-parts minimum distance, or

(e) when a magnitude of a force by which the core is supported is smaller than a previously-set prescribed threshold, or

(f) when a magnitude of a force for dropping the core is large, or

(g) when the position for supporting the core is poorly balanced.

2. The machine learning device according to claim 1, wherein

the evaluation value includes at least any of

the force by which the core is supported calculated from the position and the length of the welding part,

the force for dropping the core calculated from the position and the length of the welding part, and

balance of positions for supporting the core calculated from the position and the length of the welding part.

3. The machine learning device according to claim 1, further comprising:

a memory configured to

store the result of the machine learning, and

output the stored result of the machine learning to the processor when the processor uses the result of the machine learning section.

4. The machine learning device according to claim 1, wherein

the machine learning device is connected to at least one further machine learning device and the processor is further configured to mutually exchange or share the result of the machine learning with the at least one further machine learning device.