A Statistical Model-based Diagnostic Scheme for Cost-Effective Determination of Freshness in Fish Industry

摘要

A statistical diagnostic scheme investing on the novel concept of freshness (instead of spoilage) determination in fish industry is presented. Drawing on fault diagnosis expertise, the scheme relies on cost-effective vibration-like testing of fish samples and accurate (under typical uncertainties) stochastic modeling of their response. Via the identified models, specific quantities related to the sample's texture (itself indicative of its freshness) are computed. Then, such quantities from test samples of unknown freshness are compared with the nominal "fresh-fish" values for determining the freshness degree. This comparison is based on custom-built statistical hypothesis tests, capable of issuing reliable decisions while taking into account the risk of the decision-making process. Tests involving 0-, 3-and 6-day-old fish samples show very promising results.