Artificial neural network simulation of Ball and Stumbo formula methods ofthermal process calculations

Citation
M. Afaghi et Hs. Ramaswamy, Artificial neural network simulation of Ball and Stumbo formula methods ofthermal process calculations, J FD SCI M, 38(5), 2001, pp. 439-446
Citations number
28
Language
INGLESE
art.tipo
Article
Categorie Soggetti
Food Science/Nutrition
Journal title
JOURNAL OF FOOD SCIENCE AND TECHNOLOGY-MYSORE
ISSN journal
0022-1155 → ACNP
Volume
38
Issue
5
Year of publication
2001
Pages
439 - 446
Database
ISI
SICI code
0022-1155(200109/10)38:5<439:ANNSOB>2.0.ZU;2-S
Abstract
Application of Artificial Neural Network (ANN) models to simulate Ball and Stumbo methods of thermal process calculations is presented in this study. ANN models were developed based on relating Ball and Stumbo table parameter s to facilitate process calculations. The ANN models related g value (as a measure of process time) to f(h)/U (as a measure of process lethality). Tab les developed by Stumbo accommodate the j(infinity) (cooling lag factor) wh ile relating g and f(h)/U, therefore, for developing ANN models of Stumbo m ethod, j(infinity) was considered as an additional input variable. The deve loped ANN models for Ball and Stumbo methods were validated using a new set of processing conditions, involving a range of retort temperatures, initia l temperatures, heating rates and heating lag factors [and additionally, fo r Stumbo method cooling lag factors (j(infinity))]. The prediction efficien cy of ANN models were a function of the size of training data set, number o f hidden layers and PEs in each hidden layer as well as other learning para meters. ANN based Ball models had an average error of 1 % for the validatio n data set while the ANN based Stumbo models had a bit higher 3 % average e rror for process time and process lethality calculations. The smaller numbe r of data sets and a wider range of parameters associated with Stumbo table s were considered to be the reason for the associated higher errors with th e ANN-based Stumbo models. In general, the ANN models were considered to we ll simulate Ball and Stumbo methods of process calculations providing a bas is for further exploration of the concept.