Robot Learning and Self Optimization of Process Parameters

摘要

Robotic assembly self learning is accomplished through the use of a parameter optimization technique in the midst of running production on a real-world torque converter assembly process. The robotic performance metrics optimized in the manufacturing process are First Time Through (FTT) percentage and cycle time. The data generated is automatically gathered and analyzed by two robot program modules – RAPID motion program module and C# analyzing module. The optimization tool applies Taguchi full factorial Design of Experiments (DOE) that is running on parts during the production process. These results are subjected to automatic statistical analysis to discover the optimal parameter found based on FTT and cycle time performance. The efficacy of this method has been proved in several Ford Motor Company Powertrain assembly plants. The optimization program continues to run iteratively until no further improvement of the process is discovered or an engineering limit set on the parameter range is reached. The test results based on real world data are presented and analyzed in this paper.