Towards Real-Time In-Flight Ice Detection Systems via Computational Aeroacoustics and Machine Learning

Zhou, Beckett Yx; Gauger, Nicolas R.; Hauth, Jeremiah; Huan, Xun; Morelli, Myles; Guardone, Alberto

doi:10.2514/6.2019-3103

Cited by 7 publications

(5 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We briefly describe the ice detection problem, and the solvers and workflows for generating training data; more details can be found in [6]. Our overall goal is to create a real-time acoustics-based ice detection system using a ML model trained offline from a database of high-fidelity physics-based simulations.…”

Section: Ice Detection Problem and Simulation Workflowsmentioning

confidence: 99%

“…With visual assessment of ice nearly impossible under high rotor speeds, recent studies started exploring the use of acoustic signatures for detecting rotor blade ice [4,5]. In a recent paper [6], we proposed a novel approach towards developing a real-time in-flight ice detection system using computational aeroacoustics (CAA) and Bayesian neural networks (BNNs). In particular, the use of BNNs allowed machine learning (ML) predictions to also offer quantified uncertainty, reflecting the quality and credibility of the predicted values.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Correlation Effects in Bayesian Neural Networks for Computational Aeroacoustics Ice Detection

Hauth

Huan

Zhou

et al. 2020

AIAA Scitech 2020 Forum

View full text Add to dashboard Cite

In-flight rotor icing presents a serious problem in the operation of rotorcraft in cold climates, as complex ice shapes can significantly degrade the aerodynamic performance and handling characteristics of rotorcraft. Reliable real-time detection of ice formation is thus a critical enabling technology in improving rotorcraft safety. In this paper, we continue our previous work to explore a novel approach towards developing a real-time in-flight ice detection system using computational aeroacoustics and Bayesian neural networks (BNNs). We focus on our use of BNNs constructed from the simulated aeroacoustics dataset to enable rapid predictions of aerodynamic performance indicators together with quantified uncertainty. Specifically, we investigate the effectiveness and tradeoffs among several approximate Bayesian inference techniques for training BNNs: (Gaussian) mean-field variational inference (MFVI), full-covariance variational inference (FCVI), and Stein variational gradient descent (SVGD). We find the correlations in weight uncertainty to be low for this application and with our current BNN setup, although MFVI (which ignores correlations altogether) noticeably under-predicts the variance. SVGD is both computationally fast and captures non-Gaussian and correlation structures, appearing to be a well-suited method for the situation currently considered.

show abstract

Section: Ice Detection Problem and Simulation Workflowsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Correlation Effects in Bayesian Neural Networks for Computational Aeroacoustics Ice Detection

Hauth

Huan

Zhou

et al. 2020

AIAA Scitech 2020 Forum

View full text Add to dashboard Cite

show abstract

“…PoliMIce has uniquely been developed for predicting rotorcraft ice accretion and shedding [8,9]. It has also been utilized for the innovative design of rotorcraft acoustic ice detection technologies [10,11]. PoliMIce has also been extensively developed for the simulation and robust design optimization of thermal ice protection systems (IPS) [12,13].…”

Section: Introductionmentioning

confidence: 99%

Assessment of the PoliMIce toolkit from the 1st AIAA Ice Prediction Workshop

Morelli

Bellosta

Donizetti

et al. 2022

AIAA AVIATION 2022 Forum

Self Cite

View full text Add to dashboard Cite

This paper presents the Politecnico di Milano Icing Research Group's contribution to the 1 st AIAA Ice Prediction Workshop. A collection of two-and three-dimensional test cases to predict the collection efficiency and ice accretion are simulated using the PoliMIce ice accretion software suite. Test cases include the prediction of the collection efficiency on a three-elementairfoil and on a full-scale horizontal swept tail plane. Additionally, test cases for the simulation of ice shapes on a NACA23012 airfoil and on swept wings with varying degrees of sweep angle are assessed. The numerical predictions are evaluated and compared to high quality experimental measurements taken from the NASA Glenn Icing Research Tunnel. In general the numerical predictions compare favourably with the experimental measurements. Key droplet impingement characteristics including the collection efficiency peak and impingement limits are captured. Meanwhile distinctive ice features of rime and glaze ice regimes are depicted in the simulations. However, there remains scope for further improvement of models as highlighted by the more challenging test cases such as the three-element-airfoil and the test cases which produce glaze ice shapes with large horns.

show abstract

“…This work seeks to build upon previous experimental studies intuition of using noise for monitoring ice formation. 12 In particular, this work looks to progress from the previous study 13 through understanding the clear limitations of two-dimensional flow physics and noise and transcend into the higher-fidelity three-dimensional simulations to begin to accurately characterize the noise signals of rime and glaze ice structures. Neural networks based icing identification has been shown to be a powerful technique for in-flight ice detection.…”

Section: Introductionmentioning

confidence: 99%

Development of a Real-Time In-Flight Ice Detection System via Computational Aeroacoustics and Bayesian Neural Networks

Zhou

Gauger

Morelli

et al. 2020

AIAA Scitech 2020 Forum

View full text Add to dashboard Cite

Reliable real-time detection of ice formation is a critical enabling technology in improving rotorcraft safety. The development of a real-time in-flight ice detection system using computational aeroacoustics and Bayesian neural networks is presented and evaluated within this paper. A finite NACA0012 airfoil section undergoing sinusoidal pitching is used to represent the cyclic motion characteristic of a rotor in forward flight. Experimental ice shape measurements from the NASA Glenn Icing Research Wind Tunnel are used for validation of the numerical ice shapes. The flow-fields of the computed ice shapes are then determined using a hybrid RANS/LES approach and the acoustic noise signals of the ice shapes are predicted using the solid-surface Ffowcs Williams and Hawkings (FWH) analogy. A Bayesian neural network is then trained using the ice shapes and the performance indicators are mapped to the acoustic noise signals. This framework thus allows for the detection of ice and significantly the type of ice accreted over the pitching wing through differentiating between iced noise signals.

show abstract

Towards Real-Time In-Flight Ice Detection Systems via Computational Aeroacoustics and Machine Learning

Cited by 7 publications

References 16 publications

Correlation Effects in Bayesian Neural Networks for Computational Aeroacoustics Ice Detection

Correlation Effects in Bayesian Neural Networks for Computational Aeroacoustics Ice Detection

Assessment of the PoliMIce toolkit from the 1st AIAA Ice Prediction Workshop

Development of a Real-Time In-Flight Ice Detection System via Computational Aeroacoustics and Bayesian Neural Networks

Contact Info

Product

Resources

About