Kernel Fisher Discriminant Analysis Method Research Articles

Two-phase flow regime identification using multi-method feature extraction and explainable kernel Fisher discriminant analysis

PurposeIdentifying the flow regime is a prerequisite for accurately modeling two-phase flow. This paper aims to introduce a comprehensive data-driven workflow for flow regime identification.Design/methodology/approachA numerical two-phase flow model was validated against experimental data and was used to generate dynamic pressure signals for three different flow regimes. First, four distinct methods were used for feature extraction: discrete wavelet transform (DWT), empirical mode decomposition, power spectral density and the time series analysis method. Kernel Fisher discriminant analysis (KFDA) was used to simultaneously perform dimensionality reduction and machine learning (ML) classification for each set of features. Finally, the Shapley additive explanations (SHAP) method was applied to make the workflow explainable.FindingsThe results highlighted that the DWT + KFDA method exhibited the highest testing and training accuracy at 95.2% and 88.8%, respectively. Results also include a virtual flow regime map to facilitate the visualization of features in two dimension. Finally, SHAP analysis showed that minimum and maximum values extracted at the fourth and second signal decomposition levels of DWT are the best flow-distinguishing features.Practical implicationsThis workflow can be applied to opaque pipes fitted with pressure sensors to achieve flow assurance and automatic monitoring of two-phase flow occurring in many process industries.Originality/valueThis paper presents a novel flow regime identification method by fusing dynamic pressure measurements with ML techniques. The authors’ novel DWT + KFDA method demonstrates superior performance for flow regime identification with explainability.

Seismic response analysis and distribution prediction of source rocks in a survey of the South China Sea

Identification and prediction of high-quality source rocks is the key to obtaining new resources in the exploration area of Cenozoic basins in offshore China. We research the seismic response and distribution of hydrocarbon source rocks using seismic data, well curves, lithologic interpretation, and geochemical analysis. The target is the source rock development zone of the W Formation in a survey of the South China Sea. The results show that the seismic response of thick layer source rocks differ from surrounding rocks in the seismic profile (strong reflections with opposite polarity at the top and bottom and messy or chaotic reflections inside). Seismic reflections of interlayer source rocks have the characteristics of low frequency and continuous strong amplitude. The dominant frequency and maximum amplitude decrease as the number of mudstone layers increases. Through seismic petrophysical analysis, we have obtained three sensitive parameters of source rock in this survey: clay content, P-wave impedance, and elastic impedance. Different classification methods can realize the classification and prediction of hydrocarbon source rocks, among which the Kernel Fisher discriminant analysis method is the best. The prediction results are consistent with the geological background, geochemical information, and well curves.

A Gaussian-based kernel Fisher discriminant analysis for electronic nose data and applications in spirit and vinegar classification

Fisher discriminant analysis (FDA) is a very useful pattern recognition technique widely used in electronic nose system (e-nose). However, due to its linear characteristic, the classification problems of multi-class and high-dimensional e-nose data cannot be handled effectively. Therefore, a Gaussian-based kernel FDA (KFDA) method is proposed to solve multi-class and high-dimensional classification problems of complex samples such as food classification using e-nose. The key point of the method is how to determine the Gaussian kernel parameter. Firstly, according to distance discriminant analysis viewpoint, a desired kernel matrix adapted to Gaussian kernel function can be given successfully. Secondly, an evaluation function based on Euclidean distance is established for measuring the degree of approximation between actual kernel matrix containing an unknown Gaussian kernel parameter and the desired kernel matrix so as to get an optimal solution of the parameter, and then the actual kernel matrix can be definitely determined. Finally, the principal component analysis (PCA) for the actual kernel matrix is carried out. Meanwhile, FDA for the principal component matrix generated by PCA is also implemented in succession, and the KFDA is completed. Six kinds of Chinese spirit and six kinds of Chinese vinegar samples as two classification applications were respectively carried out accurately with the KFDA method; and the KFDA method is tested to be very simple and effective. The KFDA method may be promising for complex samples classification dataset of e-nose.

Face recognition based on gradient gabor feature and Efficient Kernel Fisher analysis

In this paper, a new Gradient Gabor (GGabor) filter is proposed to extract multi-scale and multi-orientation features to represent and classify faces. Gradient Gabor filters combine the derivative of Gaussian functions and the harmonic functions to capture the features in both spatial and frequency domains to deliver orientation and scale information. The spatial positions are encoded through using Gaussian derivatives which allow it to provide more stable information. An Efficient Kernel Fisher analysis method is proposed to find multiple subspaces based on both GGabor magnitude and phase features, which is a local kernel mapping method to capture the structure information in faces. The experiments on two face databases, FRGC version 1 and FRGC version 2, are conducted to compare performances of the Gabor and GGabor features. The experiment results show that GGabor yield a powerful tool to model faces, and the Efficient Kernel Fisher classifier can improve the efficiency of the original Kernel Fisher Discriminant analysis method.