Predicting the Crunchiness of a Chip

Predicting the Crunchiness of a Chip2019-10-24T11:41:55-04:00

Project Description

Textural properties of snack foods are essential to customer satisfaction but they are very difficult to measure on-line.

Multivariate modelling of acoustical sensor data can be used to effectively predict these important properties in real-time at low cost and without interfering with the process.

Project Details

Expertise:Machine Vision

The Challenge

Cruchiness of a Potato Chip

This snack food producer knows the importance of textural measurements such as blister level and crunchiness on mouthfeel and customer satisfaction.

Traditionally these are measured using sensory panels, meaning that only a tiny subset of the chips are analyzed and at large expense. The goal of this work was to determine the optimal way to characterize chip texture and predict this using on-line multivariate analysis of acoustical sensor data.

The Results

This study was very successful. First, the chip’s quality was quantified by combining the results of a sensory panel, visual appraisal and mechanical breaking force through a multivariate analysis.

The two calculated properties (chip brittleness and blister level) were then accurately predicted from an acoustical sensor installed on the process. Work is currently underway to commercialize this promising technology.

The ProSensus Approach

Predicting Organoleptic PropertiesIn order to be able to use acoustic sensors to predict chip textural properties, they first need to be quantified. This was accomplished by combining the results of a sensory panel, a visual appraisal, and mechanical breaking force of the chip.

A multivariate analysis accurately summarized all of this data into two variables (principal components) that are related to the chip’s brittleness and blister level.

The figure shows that summarizing all of these variables into two components could also potentially be used to specify a boundary where chips fall into an acceptable range. However, the goal is to get away from using expensive sensory panels and manual visual appraisal in order to be able to reduce costs and increase inspection rates.

An acoustical sensor was installed on the process to measure the sound of the chips on the conveyor. The operators know that the sound is related to how brittle and blistered the chips are, thus it should be possible to predict these levels from such a sensor. The acoustical signal was first pre-processed to remove background noise. Then, features from the frequency domain were used in a multivariate PLS model to predict the blister level and brittleness of the chips.

The observed versus predicted plots below clearly show that the blister level is very well predicted in both the model building data set and a test set. Results are similar for the brittleness measurement.

Organoleptic Properties Data Analysis


  1. M-J. Bruwer, J.F. MacGregor, W.M. Bourg Jr.,”Fusion of sensory and mechanical testing data to define measures of snack food texture”, Food Quality and Preference, 18, 890-900, 2007.
  2. M-J. Bruwer, J.F. MacGregor, W.M. Bourg Jr.,”Soft Sensor for Snack Food Textural Properties Using On-Line Vibrational Measurements”, Ind. Eng. Chem. Res., 46, 864-870, 2007.


Case Studies
Multi-Camera Machine Vision for Food Safety
Several of ProSensus' recent custom machine vision installations in the foods industry have included both color and thermal cameras. The digital images obtained from this combination of cameras provide numerous measurements that can be used to accomplish a variety of goals.
360 Degree Surface Inspection
One of the world's largest synthetic rubber producers came to ProSensus with customer-identified quality challenges. ProSensus designed sensitive new algorithms for the rubber inspection process and installed a 6-camera system to automatically detect defects.