A study of process optimization for roasting Taiwanese coffees by applying Six Sigma

This study involved Arabica coffees of Taiwan as the subject. The DMAIC approach in Six Sigma was introduced in the roasting process; the tools of TRIZ were employed to identify the factors of characteristics that affect the quality of coffee bean roast, where key factors were sorted out by experts. An experiment design of Taguchi method and process capability indices (Cpk) were then applied to test for the optimal parameters. The test results indicated that the process capability index increased from 0.97 before improvement to 2.69, with cup tests all proved exceeding the standards of connoisseur coffee. Last, the Back-propagation neural network was used to verify the experimental process herein on the software of Matlab R2013a version. Training results and the predicted minimal errors along with the optimal results were analyzed by mean square errors (MSE), which finally achieved convergence, whereby to determine that the optimal combinations of parameters obtained by Taguchi method can effectively stabilize the coffee roasting process and increase the quality of beans therein.