The analysis of infrared spectroscopy of substances is a non-invasive measurement technique that can be used in analytics. Although the main objective of this study is to provide a review of machine learning (ML) algorithms that have been reported for analyzing near-infrared (NIR) spectroscopy from traditional machine learning methods to deep network architectures, we also provide different NIR measurement modes, instruments, signal preprocessing methods, etc. Firstly, four different measurement modes available in NIR are reviewed, different types of NIR instruments are compared, and a summary of NIR data analysis methods is provided. Secondly, the public NIR spectroscopy datasets are briefly discussed, with links provided. Thirdly, the widely used data preprocessing and feature selection algorithms that have been reported for NIR spectroscopy are presented. Then, the majority of the traditional machine learning methods and deep network architectures that are commonly employed are covered. Finally, we conclude that developing the integration of a variety of machine learning algorithms in an efficient and lightweight manner is a significant future research direction.

Traditional algorithms cannot readily address the fact that artificial olfaction in a dynamic ambient environment requires continuous selection and execution of the optimal algorithm to detect different gases. This paper presents a deep learning WCCNN-BiLSTM-many-to-many GRU (wavelet coefficient convolutional neural network–bidirectional long short-term memory–many-to-many-gated recurrent unit) model for qualitative and quantitative artificial olfaction of gas based on the automatic extraction of time-frequency domain dynamic features and time domain steady-state features. The model consists of two submodels. One submodel recognizes a gas by the WCCNN-BiLSTM model, and the experiments based on actual data from our fabricated artificial olfactory system demonstrate that the gas recognition accuracy is nearly 100%. The other submodel quantifies the gas by the many-to-many GRU model with less labeled data; this submodel is comparable to conventional algorithms such as DT (decision tree), SVMs (support vector machines), KNN (k-nearest neighbor), RF (random forest), AdaBoost, GBDT (gradient-boosting decision tree), bagging, and ET (extra tree) according to PCA (principal component analysis) dimensionality reduction. The experimental results of 10-fold cross-validations show that the proposed many-to-many GRU outperforms the aforementioned conventional algorithms with remarkable metrics and can maintain higher concentration estimation accuracy for different unknown gases with less labeled data.

A novel gas recognition and concentration detection algorithm consisting of a dynamic wavelet convolutional neural network (DWCNN) and a many-to-many long short-term memory-recurrent neural network (LSTM-RNN), respectively, is proposed as a replacement for the traditional data processing algorithm in artificial olfaction. The proposed DWCNN gas recognition algorithm does not require gas signal preprocessing, and it directly converts the raw time-domain gas signal data to 64 * 64 2-D gray images as the input layer of a convolutional neural network (CNN). The experiments show that the recognition accuracy of CO, H ₂ , and the gas mixture of CO and H ₂ is nearly 100%, and the many-to-many LSTM-RNN algorithm requires only a few labeled data from the steady-state values of the gas sensor array signals. In addition, comparisons with other neural network multilayer perceptrons (MLPs), gated recurrent unit (GRU) algorithms, and conventional algorithms, such as the Bayesian ridge, support vector machines (SVMs), decision tree, k-nearest neighbor (KNN), random forest, AdaBoost, gradient-boosting decision tree (GBDT), and bagging, revealed that the algorithm can obtain a higher concentration detection accuracy, which was evaluated using two different kernel functions for the kernel principal component analysis (KPCA) dimensionality reduction: polynomial and rbf. The experimental results demonstrated that the proposed many-to-many LSTM gas concentration detection model outperformed the abovementioned algorithms and can more accurately estimate the concentration of different gases while using less labeled data.

Accurately and quickly identifying the gas composition and estimating the concentration are critical for ensuring industrial gas safety. However, conventional gas discrimination and concentration estimation models are unable to directly employ the raw dynamic response signal of the sensor array to accurately identify gases and estimate their concentrations online. To overcome this limitation, a cascaded approach that combines a dynamic wavelet coefficient map-axial attention network (DWCM-AAN) model for identifying gases and a prelayer normalization weighted dynamic response signal-cosformer (WDRS-cosformer) for estimating the concentration of each gas component is developed in our work. Both models directly employ the real-time dynamic response signals of the sensor array as input without any signal preprocessing. Experimental validation of CO, $\rm \textbf {H}_{2}$ , CO, and $\rm \textbf {H}_{2}$ gas mixture on our fabricated artificial olfaction revealed that the DWCM-AAN model can achieve nearly 100% accuracy in gas identification and enhance identification in real time with fewer labeled data samples. Moreover, our proposed WDRS-cosformer model achieves greater precision in concentration estimation for all different gases compared to existing state-of-the-art concentration estimation methods.

This paper studies a parameter estimation problem for the non-linear diffusion equation within multiphase porous media flow, which has important applications in the field of oil reservoir simulation. First, the given problem is transformed into an optimization problem by using optimal control framework and the constraints such as well logs, which can restrain noise and improve the quality of inversion, are introduced. Then we propose the widely convergent homotopy method, which makes natural use of constraints and incorporates Tikhonov regularization. The effectiveness of the proposed approach is demonstrated on illustrative examples.

Conventional gas recognition algorithms used in artificial olfaction rely on manually selected steady-state features of the sensor array signal and ignore the dynamic process signal response features of the sensor array; hence, a large amount of dynamic feature information, which can improve the gas recognition accuracy, is lost. To address this problem, a wavelet transform coefficient map-capsule network (WTCM-CapsNet) model for gas recognition is presented in this paper. It directly uses a raw signal from a sensor array without any signal-processing steps and consists of a wavelet coefficient map and a capsule network (CapsNet) model with two submodels. The WTCM-CapsNet model directly performs a five-scale wavelet transform on the sensor array signal and extracts 4,096 wavelet coefficients to form a 2D ${64}^{ \boldsymbol {\ast }}{64}$ wavelet coefficient image as the CapsNet input. Compared with conventional gas recognition algorithms, such as the convolutional neural network (CNN) with 99.3% gas recognition accuracy, the K-nearest neighbor (KNN) with 95.74% gas recognition accuracy, support vector machine (SVM) with 96.45% gas recognition accuracy, random forest with 95.74% gas recognition accuracy, naive Bayes with 92.91% gas recognition accuracy and BP neural network (BPNN) algorithms with 97.87% gas recognition accuracy, which extract the steady-state values of the time-domain signal and the wavelet coefficient maps of the whole dynamic response signal in the time-frequency domain as input features, CapsNet uses an iterative routing-by-agreement mechanism that can effectively improve the gas recognition accuracy to approximately 100%, as demonstrated by the dynamic response signal of the gas sensor array collected by our fabricated artificial olfactory system.