PCA and machine learning analysis of signals to measure scalars in reacting flows

Laser-induced grating spectroscopy (LIGS) is a technique used to measure the local speed of sound in high temperature gas mixtures. The principle is to generate a local, instantaneous (ns) density gradient using a high power laser. The sound wave generated in the process propagates at the speed of sound, and the resulting unsteady pattern is detected using a continuous low power laser beam, which is reflected by the grating, and contains a signature of the pattern.

The signal frequency is proportional to the speed of sound, and is easily detected. However, the signal is a combination of electrostrictive and thermal waves, and it can be difficult to discern which one is dominant; this plays a role in automating the process of data analysis.

We have created new algorithms involving principal component analysis (PCA) and machine learning for analysing these signals, and performed some preliminary testing using a turbulent hydrogen diffusion flame as shown in Fig. 1. Chaib et al. (2024), J. Eng. Gas Turb. Power 146,111020-1]. We would like to deploy it to a larger dataset for turbulent premixed flames, and analyse the performance of the same code with different operating conditions, as shown in Fig. 2.

Requirements: interest in signal processing and algorithms, good proficiency with Matlab or python .