Soft Sensor for Snack Food Textural Properties Using On-Line Vibrational Measurements

Abstract

It is common in the food industry to produce solid-phase products that are not amenable to ready characterization on the production line via traditional on-line instrumentation. This motivates research into novel sensor technologies. In this work, an on-line vibrational sensor is used to develop a soft sensor for real-time prediction of product texture in a commercial snack food process. This in turn enables real-time multivariate statistical process control (MSPC) and indicates the opportunity for automated feedback control. An accelerometer is used to record the vibrational signature generated by a snack food product falling onto a metal surface. The frequency distribution of the acoustic signature is obtained via the discrete Fourier transform. This provides a predictor space from which the textural properties are modeled using partial least-squares (PLS). Excellent results are obtained, with R2 upward of 96% on an independent test set for two properties that fully span the texture space of this product.