This paper considers the potential for identifying industrial manufacturing conditions that will lead to high acrylamide formation in potato crisp manufacture. Considering the available historical industrial processing data, initial tests were undertaken to identify the degree of variability and confidence in the data. Following data visualisation which indicated data 'fingerprints' characteristic of high acrylamide, Partial Least Squares (PLS) Discriminant Analysis (DA) was implemented to provide indications of the probability that high acrylamide product would be produced. It was determined that in a third of instances, high acrylamide could be predicted while maintaining a low level of false predictions. The predominance of fructose concentration in the prediction along with the need for asparagine were indicated and aligned well with prior literature mechanistic model indications. The ability to identify a third of high acrylamide occurrences provides the process operators with a good opportunity to make process modifications that would comply with increasingly stringent regulation.
Bibliographical noteFunding Information:
This work was supported by financial assistance of InnovateUK through the MCAP2 - Measurement and Control of Acrylamide in Production Processes project (Ref 103946 ).
© 2020 Elsevier Ltd
Copyright 2020 Elsevier B.V., All rights reserved.