Abstract: In this work, we optimized the frying process for sweet potato chips using combination of one-factor-at-a-time method and uniform design through multiple regression analysis of the uniform design experimental data. The results of the one-factor-at-a-time experiments showed slice thickness, pre-drying and frying to be the factors with a great influences on the oil content, L* and b*. Uniform design and stepwise regression analysis indicated that frying time significantly influenced the oil content (P = 0.011), saline concentration influenced significantly water content (P = 0.022), frying temperature and frying time affected very significantly L* (P = 0.001), drying time and frying time significantly influenced b*, and saline immersion time, saline concentration, pre-drying temperature and time and slice thickness affected brittleness, which was fitted to a quadratic polynomial regression model (P = 0.001). Principal component analysis (PCA) showed that the two principal components extracted accounted for 87.4% of the total variation, thus achieving dimensional reduction. Through ridge regression analysis, we established a regression model for comprehensive score with a correlation coefficient R of 0.997, suggesting a good degree of fitting. Partial least squares regression analysis predicted that the best process parameters were as follows: slice thickness, 2 mm; blanching time, 1 min; saline immersion time, 20 min; saline concentration, 1%; pre-drying temperature, 60 ℃; pre-drying time, 70 min; frying temperature, 150 ℃; and frying time, 1 min. The validation experiment gave a comprehensive score of 0.89, which was higher than the highest score (0.86) from uniform design. The product obtained under the optimized conditions had a low oil content, and accepted color and brittleness, and the proposed model had good prediction ability.

Key words: uniform design, principal component analysis, comprehensive assessment, oil content, color, brittleness