An a posteriori data-driven method for phase-averaged optical measurements

E. Amico, S. Montagner, J. Serpieri, G. Cafiero · 2025

Read An a posteriori data-driven method for phase-averaged optical measurements by E. Amico, S. Montagner, J. Serpieri, G. Cafiero (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12369.txt.

Introduction

Enrico Amico, Sara Montagner, Jacopo Serpieri, and Gioacchino Cafiero recover phase-averaged flow fields from random snapshots, with no synchronized camera rig. Their paper from twenty twenty-five serves fluid experiments on propellers, turbines, and pulsed jets. Classic phase-averaging demands costly high-speed gear and precise triggers. Their ISOMAP-based method instead reorders uncorrelated frames after the fact, placing each in phase space.

Simulations and real optical datasets, with single and multi-propeller setups, validate the approach. The reordered frames yield phase-averaged fields that capture the periodic flow well. Sync hardware limits vanish from the budget. Archival photo sets, once unusable for phase work, become fair game. The team pitches the method as a scalable, cheap stand-in for rigged acquisition in periodic flow studies.

Catalog record

CREATORE. Amico, S. Montagner, J. Serpieri, G. Cafiero
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-12369

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

An a posteriori data-driven method for phase-averaged optical measurements E. Amico, S. Montagner, J. Serpieri, G. Cafiero Abstract Phase-averaging is a fundamental approach for investigating periodic and non-stationary phenomena. In fluid dynamics, these can be generated by rotating blades such as propellers/turbines or by pulsed jets. Traditional phase-averaging approaches often rely on synchronized data acquisition systems, which might require high-speed cameras, light sources, and precise delay generators and encoders, making them expensive and sometimes unfeasible. This work proposes an a posteriori data-driven approach that reconstructs phase information from randomly acquired uncorrelated photographic frames (snapshots) using the ISOMAP algorithm. The technique enables accurate reordering of snapshots in the phase space and subsequent computation of the phase-averaged flow field without the need for synchronization. The framework was validated through numerical simulations and experimental fluid dynamics datasets from an optical setup featuring single- and multi-propeller configurations. The results demonstrate that the proposed method effectively captures the periodic flow characteristics while addressing the challenges related to synchronization and hardware limitations. Furthermore, the ability to apply this technique to archival datasets extends its applicability to a wide range of experimental fluid dynamics studies. This approach provides a scalable and cost-effective alternative to traditional methods for the analysis of periodic phenomena. 1 Introduction Investigating non-stationary and periodic phenomena presents notable challenges due to their different nature: While some phenomena are directly correlated with the underlying periodic mechanisms at play, these might cause/influence a cascade of effects with a much broader spectral content. An example of this occurs in fluid dynamic, especially in contexts like the wake generated by propellers/turbines or flows generated by a periodic actuation, for instance, zero-net-mass-flux jets. Such flows exhibit periodic and organized patterns, making their accurate analysis reliant on specialized techniques. One widely used method for examining these flows involves performing phase-averaged measurements/simulations, which focus on isolating the periodic characteristics by aligning data acquisition with the flow’s inherent periodicity. Phase-averaging techniques have been widely employed to investigate flow fields in such contexts. For example, Greco et al. employed a methodology for phase averaging by synchronizing a laser with the piston motion in impinging jet experiments. This approach utilized high-speed lasers and cameras, ensuring precise timing and detailed capture of flow structures at various phases of the cycle. Synchronization enabled the acquisition of phase-resolved data, which is critical to understanding the periodic behavior of the flow. Similarly, Cafiero et al. studied the evolution of the starting vortex generated by a synthetic jet when interacting with a square fractal grid. Examples of applications of phase-averaged measurements can be found in the fields of combustion [ 20 ] , the evolution of propeller wakes [ 10 , 16 ] or of boundary layer instability mechanisms [ 19 ] . In the case of propeller-induced flows, phase-averaged flowfield studies are particularly valuable for analyzing the periodic and organized variations in the flow surrounding the blades. Such analyses are pivotal for applications in distributed propulsion systems and drones for urban air mobility (UAM), where propeller interactions significantly influence efficiency and noise characteristics. However, conducting phase-averaged measurements in experimental setups presents several challenges. In the case of propellers, for instance, maintaining a stable rotational speed can be particularly cumbersome, as it requires high-quality motors and sophisticated controllers, which can be difficult to implement, especially with the smaller motors used for small drones’ propellers. Alternatively, capturing time-resolved flowfield data often involves the deployment of high-speed velocimetry setups with expensive lasers, cameras, and synchronizing systems. To address these challenges, this article proposes an alternative method for phase-averaged analysis that does not rely on synchronized data acquisition. Instead, it utilizes an a posteriori phase reconstruction from randomly acquired, uncorrelated data. This approach is particularly advantageous when advanced equipment is not available. Additionally, the method can be applied to previously collected datasets, extending its utility to archival data. By isolating phase information after data acquisition, this methodology offers a flexible and cost-effective solution to study periodic phenomena, expanding the accessibility to phase-averaged techniques in experimental measurements. 2 Methodology 2.1 Phase averaging of optical measurements A n -dimensional manifold is a topological space in which every point has a neighborhood homeomorphic to an Euclidean space \mathbb{R}^{n} . The proposed approach employs the ISOMAP algorithm [ 21 ] to discern a low-dimensional embedding. One notable application is in mass cytometry data analysis, where ISOMAP improves the separation of different cell types, leading to enhanced classification accuracy [ 4 ] . In hydrogeology, it has been applied for high-resolution characterization of groundwater systems, using well log data to optimize regression parameters [ 12 ] . In industrial diagnostics, ISOMAP, in combination with machine learning techniques, has been successfully employed for fault diagnosis in high-speed train air compressors, improving anomaly detection accuracy [ 14 ] . Furthermore, in the field of brain-computer interfaces (BCI), a modified version of ISOMAP, known as Supervised Discriminant ISOMAP Projection (SD-ISOMAP), has been utilized to enhance EEG signal classification, making BCI systems more efficient and practical for real-time applications [ 17 ] . Additionally, ISOMAP has found applications in digital marketing, where it has been used in conjunction with big data technologies to improve sentiment analysis and emotion-based consumer behavior studies on e-commerce platforms [ 25 ] . ISOMAP has also been successfully implemented in manifold learning [ 5 ] .These examples highlight the versatility of ISOMAP in diverse domains. (a) Figure 1: Schematic representation of the phase non-uniqueness problem observed from a plane normal to the plane of the propeller. The grey area represents the camera’s viewing angles. The dashed lines represent the projection of the blade tip onto the camera plane. (a) Figure 2: Schematic representation of the framework developed and employed for the phase reconstruction starting from the data. Particle image velocimetry (PIV) involves acquiring for each vector field a pair of particles’ images with a known time delay, \Delta_{T} . The value of \Delta_{T} is selected to ensure particle correlation, generally leading to a displacement between the two snapshots of about 10-12 pixels [ 23 , 15 ] . Stereoscopic PIV (S-PIV) requires two independent views of the region of interest, thus allowing the measurement of the out-of-plane component of the velocity field [ 24 ] through geometric corrections. Therefore, in the S-PIV case, each vector field requires two pairs of snapshots, one pair per camera. The purpose of the present study is to investigate the wake generated by a propeller operating under different inflow conditions. Since the study is primarily focused on the flowfield in the immediate vicinity of the blade, the propeller is imaged within the field of view (FOV). The plane considered is perpendicular to the propeller’s rotation plane and intersects its axis of rotation. Considering the expected velocities within the field of view [ 7 ] and the optical setup features, \Delta_{T} is on the order of tens of \mu\,\mathrm{s} ; on the other hand, the characteristic time of the propeller, T_{p} , i.e. the time required to complete one revolution, is on the order of 10^{-3}\,\mathrm{s} , considering rotational speeds of the propeller ranging from 1000 to 5000 rpm . Consequently, two snapshots belonging to the same pair contain information about the propeller’s rotational motion. Therefore, the particles’ images include both the

Read the complete text