Point Relationship Research Articles

Data-driven approach for the classification of gas turbine faults

Gas turbines (GTs) play a crucial role in the production of electricity. Extreme working conditions can lead to deterioration in GTs' performance, resulting in the occurrence of various issues. This study proposes an approach to deal with this issue by combining a layered recurrent neural network (LRNN) with principal component analysis (PCA). This approach aims to reduce the dimensionality of data and computational complexity effectively while enhancing the accuracy of gas turbine fault classification. The methodology outlined consists of two steps. The first step is to apply PCA to the dataset that was collected from the gas turbine. By transforming the data into a lower-dimensional space, PCA helps to eliminate redundant information and improve computational efficiency. Next, LRNN is employed to detect and classify faults in the gas turbine. The LRNN’s structure enables it to capture complex patterns and relationships in the data, which enhances the accuracy of fault classification. This study is based on historical data collected from a gas turbine power station, consisting of 8200 samples of 34 measured variables. The operating parameters contain data such as temperature and pressure. Each data point's relationship to a specific turbine scenario reveals if it is healthy or one of the four faulty scenarios. The results showed that by combining the LRNN with PCA, the gas turbine fault classification achieved good performance in terms of accuracy, precision and neural network model performances, while also showcasing the faster convergence speed of the LRNN when trained on PCA instead of the original dataset.

Stratifying High-Dimensional Data Based on Proximity to the Convex Hull Boundary

The convex hull of a set of points, $C$, serves to expose extremal properties of $C$ and can help identify elements in $C$ of high interest. For many problems, particularly in the presence of noise, the true vertex set (and facets) may be difficult to determine. One solution is to expand the list of high interest candidates to points lying near the boundary of the convex hull. We propose a quadratic program for the purpose of stratifying points in a data cloud based on proximity to the boundary of the convex hull. For each data point, a quadratic program is solved to determine an associated weight vector. We show that the weight vector encodes geometric information concerning the point's relationship to the boundary of the convex hull. The computation of the weight vectors can be carried out in parallel, and for a fixed number of points and fixed neighborhood size, the overall computational complexity of the algorithm grows linearly with dimension. As a consequence, meaningful computations can be complete...