| CPC B23K 26/40 (2013.01) [B23K 26/382 (2015.10); G06F 17/18 (2013.01); G06N 20/00 (2019.01); B23K 26/0622 (2015.10)] | 13 Claims |
|
1. A computer-implemented method for operating a laser material processing machine, the method comprising:
beginning with an initial set of respective values as respective current values of a plurality of process parameters, performing an iterative process that gradually modifies the respective current values to a final set of the respective values, the iterative process including the following in each iteration of the iterative process, except a final one of the iterations:
(a) actuating, by a processor system that includes at least one processor, the laser material processing machine to perform a machining of a workpiece while set with the respective current values of the plurality of process parameters;
(b) obtaining sensor data by a sensor system sensing results of the machining performed in the respective iteration;
(c) determining, by the processor system and from the obtained sensor data, a respective quality value for each of a plurality of quality parameters;
(d) determining, by the processor system, a cost of the current values of the respective iteration based on a combination of respective scaled differences of the quality values from respective target values of the respective quality parameters;
(e) determining, by the processor system, whether the cost exceeds a threshold cost; and
(f) in response to the cost being determined to exceed the threshold cost, modifying one or more of the respective current values of the plurality of process parameters, the current values thereby being updated for use in an immediately following other one of the iterations;
wherein:
the iterative process continues with the immediately following other of the iterations as long as the cost exceeds the cost threshold, and otherwise terminates with the current values of a last of the iterations being the final set, the final one of the iterations including steps (a)-(e); and
the modifying the one or more of the respective current values of the process parameters is performed using Bayesian optimization that is performed (i) with a data-based process model, and (ii) to thereby select which of process parameters' respective current values are to be modified and the way in which the respective current values of the selected process parameters are to be modified based on maximizing a probability that respective variables of each of the plurality of quality parameters will be within respective predefined low and high boundaries.
|