Files
Abstract
Conventional snack food texture analysis is costly. This research aimed to develop an affordable snack food texture analysis technique. Two tests assessed chips using sensory, mechanical, and acoustical evaluations. The first test involved four chips: LAY'S® Classic, LAY'S® Kettle Cooked, RUFFLES® Original, and DORITOS® Nacho Cheese. The second test modified LAY'S® Classic Chips with varying water activities (0.17, 0.25, 0.35, 0.45). Both studies had the chips undergo a modified Quantitative Descriptive Analysis, where nine panelists rated “crispness”, “crunchiness”, and “crackliness”. Instruments measured mechanical and acoustical factors related to “crispness”, like force and sound pressure. Sensory and instrumental measurements revealed significant differences between chip types. In both tests, machine learning models were built that classified chips by breaking sound. These models were then duplicated and labeled to instead predict “crispness”, “crunchiness”, or “crackliness” from the breaking sound, totaling eight models. The models had an accuracy of between 87.58% and 96.58%.