In this study, a portable electronic nose (E-nose) prototype is developed using metal oxide semiconductor (MOS) sensors to detect odors of different wines. Odor detection facilitates the distinction of wines with different properties, including areas of production, vintage years, fermentation processes, and varietals. Four popular machine learning algorithms—extreme gradient boosting (XGBoost), random forest (RF), support vector machine (SVM), and backpropagation neural network (BPNN)—were used to build identification models for different classification tasks. Experimental results show that BPNN achieved the best performance, with accuracies of 94% and 92.5% in identifying production areas and varietals, respectively; and SVM achieved the best performance in identifying vintages and fermentation processes, with accuracies of 67.3% and 60.5%, respectively. Results demonstrate the effectiveness of the developed E-nose, which could be used to distinguish different wines based on their properties following selection of an optimal algorithm.

Electronic nose, as a non-destructive instrument, is widely used in the field of gas analysis. In this work, E-nose was employed to distinguish wines and Chinese liquors by means of a machine learning technique. First, a multi-hidden layers Back-Propagation Neural Network (BPNN) was designed to build an identification model for the classification of different wines. Then, a BPNN-based transfer-learning framework was developed with minimal changes to the architecture of the BPNN-based model which was trained on the wine sample dataset. Experimental results revealed that the BPNN-based model performed with a 98.27% accuracy in identifying different wines, and the BPNN-based transfer-learning framework performed with a 93.4% accuracy in identifying Chinese liquors by only retraining the output layer. This reduced the model training costs compared with the complete retraining of a new classification model. Results demonstrated the effectiveness of the proposed BPNN-based transfer-learning model, which was capable of identifying different kinds of wines based on their own properties and could be easily applied to the classification of Chinese liquors. The model-based transfer learning framework offered promising potential for different classification tasks of various beverage.

The quality grades of organic green teas are tightly correlated with their prices. In this work, samples of organic green teas of different quality grades are collected, and their aromas are analyzed with an electronic nose (E-nose). A multi-task model based on the back propagation neural network (MBPNN) is proposed for the simultaneous performance of the classification task (grade classification of tea) and regression task (quality prediction of tea with market price). The validity of the proposed MBPNN model is also verified; its performances of the tasks are compared with those of two classification models (random forest and support vector machine) and three regression models (partial least squares regression, kernel ridge regression, and support vector regression). Experimental results demonstrate that the MBPNN model achieves good performance both in the tasks of tea grade classification and tea quality evaluation (price regression). The study shows that the E-nose is effective for the classification and evaluation of organic green teas when an optimal pattern recognition algorithm is selected. Encouragingly, a novel application of the multi-task learning model in the tea industry is obtained to assess the tea quality in a simple, quick, and comprehensive way. INDEX TERMS Classification model, electronic nose, multi-task framework, organic green teas, regression model.

Remaining useful life (RUL) estimation is fundamental to prediction and health management (PHM) technology. Traditional machine learning generally assumes that the training and testing sets are independent and identically distributed. As distribution differences exist in real scenarios, this assumption hinders the effectiveness of the traditional machine learning methods. Aiming at these problems, we propose a CNN-LSTM-based domain adaptation framework for RUL prediction in this work. A shared encoding network and domain adaptation mechanism is introduced to decrease the data distribution discrepancy between the source and target domains. A cross-linking architecture is also developed for feature fusion, which considers the features at different levels to guarantee that the generated fusion features contain sufficient information for prognosis. Extensive experiments are then conducted to verify the superiority of the proposed framework. The experimental results demonstrate that the proposed model has excellent performance, especially for equipment with more complex working conditions and data.