Supporting Materials for "Airflow Velocity Estimation from a Fluttering Tuft"

Description of Supporting Materials

This repository provides supplementary figures, a supporting video, and numerical data associated with the study "Airflow Velocity Estimation from a Fluttering Tuft." The materials provide additional information on neural-network architecture comparisons, long-term measurement stability, and the statistical characteristics of the datasets used for model development and evaluation.

The repository contains the following files:

Supplementary figures

• Fig. R1. Performance comparison of capacity-matched LSTM architectures.png: Compares the velocity estimation performance of LSTM architectures with different numbers of recurrent layers and an attention-based LSTM, using approximately matched numbers of trainable parameters.

• Fig. R2. MFC feedback signals for tuft trials.png: Presents the mass flow controller (MFC) feedback signals recorded during two long-term tuft measurement trials.

• Fig. R3. Evolution of signed percentage errors of tuft trials.png: Shows the temporal evolution of signed percentage errors in the tuft-based velocity estimates during the long-term measurement trials, providing information on estimation drift.

Supplementary tables

• Table RI Architectural configurations of capacity-matched LSTM models.csv: Lists the architectural configurations of the compared LSTM models, including the number of recurrent layers, hidden units, attention dimensions, trainable parameters, and relative differences in parameter counts.

• Table RII Statistics of input displacement and reference velocity.csv: Provides descriptive statistics (mean, standard deviation, minimum, and maximum) of the three-dimensional tuft-tip displacement components (x, y, and z) and reference airflow velocity for the training, validation, and testing datasets.

Supporting video

• Supporting video.mp4: Provides a visual demonstration accompanying the proposed tuft-based airflow sensing method.

These materials are made publicly available to facilitate the interpretation of the reported results and support further analysis of the proposed airflow sensing method.