The improvement of quality management has always been one of the main KPIs of manufacturers. Cyntec Co., Ltd. realizes that the quality related research as well as the collection and tracking of production history are inseparable from AI technology. Data collection with in-depth data analysis can improve quality rapidly. In light of this knowledge, Cyntec has invested in a team of experts in manufacturing and has invited DRC to conduct data analysis and diagnostics for a joint development in verification solutions.

The early stages of this project also faced challenges such as “new equipment parameter tunings for weeks”, “highly complex and mutually interfering parameters obscures the correlation between the parameters”, and “traditional experiment designs are time-consuming and makes it difficult to produce a good yield at the same time”. Thus, the two teams find the key problem through data quality diagnostics, root cause analysis on quality, and the tune the parameters to improve the product quality for first-of-a-kind (FOAK) parameter optimization.

In contrast to the prior methods of trial and error or the comprehensive planning of Taguchi Methods in which the levels of various factors are laid out, the popular algorithm Bayesian Optimization was adopted this time, combined with expert production process knowledge, to design methods through human-machine interaction mechanisms in sequential experiments. In the past, a production line would require a full-time R&D staff to tune parameters for one week. Now that period has been reduced to 1.5 hours for tuning 11 parameters in winding station line changes. The parameters recommended by the algorithm can result in 96% yield for 100 products. It is a testament to the infinite potential of AI in smart manufacturing.

Cyntec and the DRC team find the key problem through data quality diagnostics, root cause analysis on quality, and then tune the parameters to improve the product quality for project parameter optimization