Non-destructive digital technologies for quality assessment and authentication of wine and beer
Loading...
Citations
Altmetric:
Altmetric::
Download
Associated Records
Authors
File Licence
Editors
Date
2025
Source Title
Journal Title
Journal ISSN
Volume Title
Publisher
Melbourne Authors
Affiliation
Minerva Elements
Document Type
PhD thesis
Subject Code
Citation
Access Status
This item is embargoed and will be available on 2027-10-07. This item is currently available to University of Melbourne staff and students only, login required.
DOI
Access URL
Open Access URL
Open Access at PMC
Linked Resource URL
NHMRC Grant Code
ARC Grant Code
Description
© 2025 Natalie Maree Harris
Abstract
Assessment of quality and authenticity for wine and beer are traditionally done with complex chemical analysis and sensory evaluation; however, these methods are costly, time consuming, require highly specialized equipment and highly trained personnel. With the increased occurrence and economic impact worldwide with fraud there is a need for more rapid assessment approaches. Advancements in digital technologies offer non-destructive, accurate, low-cost, portable devices that may be used during winemaking, beer production, bottling process, post-bottling, storage, and transport. The digital technologies used in this research include near-infrared (NIR) spectroscopy, electronic nose (e-nose), and machine learning (ML) modeling using artificial neural networks (ANN).
Rapid liquid quality assessment for thirteen Shiraz vertical vintages was explored using digital technologies. Chemical analysis and sensory evaluation were undertaken with gas chromatography - mass spectroscopy (GC-MS), and quantitative descriptive analysis (QDA), and these were used as targets for ANN models. A hand-held portable NIR spectroscopic device (1596 – 2396 nm) obtained a chemical fingerprint for the wines that was used as inputs for ANN models, also with values from a low-cost e-nose. The eight models developed showed high accuracy for classification of vintages (NIR: 98.3%; e-nose: 99.5%), and prediction of volatile aromatic compounds (NIR: R = 0.96; e-nose: R = 0.96), sensory descriptors (NIR: R = 0.97; e-nose: R = 0.97), and physicochemical parameters (NIR: R = 0.94; e-nose: R = 0.93). These models may be used by winemakers to monitor and assess vintages to maintain high quality or style associated with wine variety, vineyard, or region.
The novel approach of using ANN and NIR spectroscopy through bottle (NIRB) was investigated as a rapid and accurate screening approach. The chemical fingerprint for the thirteen Shiraz vertical vintages was obtained without opening the wine bottle. These NIR values were used as inputs for modelling along with targets provided from GC-MS analysis and QDA sensory evaluation. The ANN models showed high accuracy for classification of 13 vintages (97.2%), and for prediction of sensory descriptors (R = 0.95), and volatile aromatic compounds (R = 0.88). These preliminary models may be developed further with different varietals, region, production and winemaking practices.
The rapid screening methods developed in previous studies were applied to six Chardonnay vertical vintages from another wine region. For rapid liquid analysis the ANN models using e-nose values as inputs, while unopened wine bottles were screened using NIRB. The eight models achieved high accuracy for classification of vintages (e-nose: 99.4%; NIRB: 99.4%), prediction of sensory descriptors (e-nose: R = 0.98; NIRB: R = 0.95), volatile aromatic compounds (e-nose: R = 0.99; NIRB: R = 0.92), and physicochemical parameters (e-nose: R = 0.95; NIRB: R = 0.96). The rapid liquid screening models may be used by winemakers during production and fermentation to give sensory and aroma profile of that vintage. The through the unopened bottle screening may be used to authenticate vintages, monitor sensory and aroma attributes during storage.
The novel approach of ANN and NIRB was used for commercial beers with different styles and fermentation types. There were three highly accurate ANN models for classifying fermentation type (99%), prediction of sensory descriptors (R = 0.92), and volatile aromatic compounds (R = 0.94). Rapid screening without the need to open the wine or beer bottle may be used to assess quality traits post bottling, during storage, transport, and point of retail to detect mislabeling and fraud, change in quality traits, and development of off-aromas and flavors.
Notes
Terms of Use
Keywords
Quality and authentication of wine and beer, Spectroscopy through bottle, Electronic nose, Machine Learning Modelling