Smart Detection of Faults in Beers Using Near-Infrared Spectroscopy, a Low-Cost Electronic Nose and Artificial Intelligence

Abstract

Early detection of beer faults is an important assessment in the brewing process to secure a high-quality product and consumer acceptability. This study proposed an integrated AI system for smart detection of beer faults based on the comparison of near-infrared spectroscopy (NIR) and a newly developed electronic nose (e-nose) using machine learning modelling. For these purposes, a commercial larger beer was used as a base prototype, which was spiked with 18 common beer faults plus the control aroma. The 19 aroma profiles were used as targets for classification machine learning (ML) modelling. Four different ML models were developed; Models 1 (M1) and M2 based on NIR (100 inputs from 1596–2396 nm) and M3 and M4 based on the e-nose (nine sensor readings as inputs) and 19 aroma profiles as targets for all models. A customized code tested 17 artificial neural network (ANN) algorithms automatically testing performance and neuron trimming. Results showed that the Bayesian regularization algorithm was the most adequate for classification rendering precisions of M1 = 98.9%, M2 = 98.3%, M3 = 96.8%, and M4 = 96.2% without statistical signs of under- or overfitting. The proposed system can be added to robotic pourers and the brewing process at low cost, which can benefit craft and larger brewing companies.

Digital Assessment and Classification of Wine Faults Using a Low-Cost Electronic Nose, Near-Infrared Spectroscopy and Machine Learning Modelling

“…Six classification ML models for each type of wine were developed using artificial neural networks. As described in previous publications from Gonzalez Viejo et al [ 15 , 33 ], a customized Matlab ® R2021a code developed by the DAFW-UoM group was used to train the models using 17 supervised algorithms classified as two algorithms based on backpropagation with Jacobian derivatives, 11 based on backpropagation with gradient derivatives and four using weight/bias training functions. The best models were selected based on the highest accuracy and performance calculated using means squared error (MSE) to assess the absence of overfitting.…”

“…Implementing these digital technologies was the first attempt to transform the grape and wine industry production from traditionally reactive to a more predictive process using smart decision making [ 13 , 14 ]. For fault detection in fermented beverages, previous research has shown high accuracy for artificial intelligence (AI) tools using NIR, e-noses, and machine learning for beer [ 15 ]. Specifically, Gonzalez Viejo et al [ 15 ] developed a method to detect 18 different faults in beer using a low-cost e-nose coupled with artificial neural networks; this technique is able to predict the level of concentration of faults and the specific compounds.…”

Evolutionary Optimisation of Beer Organoleptic Properties: A Simulation Framework

“…In related research by Rodman et al [15], the aim has been to gain insight into the brewing process and provide an investigation into the influence of byproduct (diacetyl and ethyl acetate) threshold levels on obtainable fermentation performance by computing optimal operating temperature profiles for a range of constraint levels on byproduct concentrations in the final product. Some recent research has also been exploring faults detection in beers using artificial intelligence methods, as well as using strain development methodology to breed industrial brewing yeast [16,17].…”

Computer Vision and Machine Learning Analysis of Commercial Rice Grains: A Potential Digital Approach for Consumer Perception Studies

“…This method uses independent sets of samples for each stage and evaluates the overall accuracy by including all samples. A similar data division to develop ANN models was used in previous studies [35,59,74]. Besides, several retraining attempts were conducted to assess the consistency of the results, obtaining similar results in every attempt.…”

“…The artificial neural networks (ANNs) for supervised ML are well-known for solving multiclass classifications due to their ability to deal with non-linear data for pattern recognition to obtain high accuracy. For example, the ANN models were used in previous studies to classify mulberry fruit according to the ripeness levels [27], detect beer faults using the electronic nose [35], and classify aphid infestation levels using the electronic nose and near-infrared spectroscopy [36].…”