Training bioinspired sensors to classify flows

Alsalman, Mohamad; Colvert, Brendan; Kanso, Eva

doi:10.1088/1748-3190/aaef1d

Cited by 27 publications

(33 citation statements)

References 53 publications

(76 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2019), flow field reconstruction and prediction (Maulik & San 2017; Maulik et al. 2018; Fukami, Fukagata & Taira 2019; Huang, Liu & Cai 2019; Lee & You 2019) and, more relevant to the present study, flow field identification (Colvert, Alsalman & Kanso 2018; Alsalman, Colvert & Kanso 2019; Wu et al. 2019 b ).…”

Section: Introductionmentioning

confidence: 82%

“…Once the training process is finished, the classifier can be used to classify a given sample automatically. An example of the application of the classification method in fluid mechanics is the classifier of wake pattern in the flow past an airfoil by Colvert et al (2018) and Alsalman et al (2019). They used a neural network to train the classifier for classifying the wake pattern generated by different motions of the airfoil.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Using machine learning to detect the turbulent region in flow past a circular cylinder

Yang

Zhang

et al. 2020

J. Fluid Mech.

View full text Add to dashboard Cite

show abstract

Section: Introductionmentioning

confidence: 82%

Section: Introductionmentioning

confidence: 99%

Using machine learning to detect the turbulent region in flow past a circular cylinder

Yang

Zhang

et al. 2020

J. Fluid Mech.

View full text Add to dashboard Cite

show abstract

“…6), we find that the novel y component is more informative for discerning between objects. Combining the two measuring dimensions improves classification even further, which also holds for object localization [11], [36].…”

Section: D Sensingmentioning

confidence: 90%

“…More directly, zero crossings, maxima, and minima of the velocity profile can also be directly used to produce estimates for the distance (y-coordinate) and the lateral position (x-coordinate) [10]. In [30], [36], [37], it is shown that an ALL comprising of 2D-sensitive sensors makes localization more robust. Compared to the 1D sensitivity of fish and most other ALL implementations, this extension provides more and complementary spatial reference points [30], [36], which may also be helpful for shape recognition.…”

Section: ) Velocity Profilesmentioning

confidence: 99%

Shape Classification Using Hydrodynamic Detection via a Sparse Large-Scale 2D-Sensitive Artificial Lateral Line

et al. 2020

View full text Add to dashboard Cite

Artificial lateral lines are fluid flow sensor arrays, bio-inspired by the fish lateral line organ, that measure a local hydrodynamic environment. These arrays are used to detect objects in water, without relying on light, sound, or on an active beacon. This passive sensing method, called hydrodynamic imaging, is complementary to sonar and vision systems and is suitable for collision avoidance and near-field covert sensing. This sensing method has so far been demonstrated on a biological scale from several to tens of centimeters. Here, we present measurements using a large-scale artificial lateral line of 3.5 meters, consisting of eight all-optical 2D-sensitive flow sensors. We measure the fluid flow as produced by the motion of five different objects, towed across a swimming pool. This results in repeatable stimuli, whose measurements demonstrate a complementary aspect of 2D-sensing. These measurements are both used for constructing temporal hydrodynamic signatures, which reflect the object's shape, and for flow-feature based near-field object classification. For the latter, we present a location-invariant feature extraction method which, using an Extreme Learning Machine neural network, results in a classification F1-score up to 98.6% with selected flow features. We find that, compared to the traditional sensing dimension parallel to the sensor array, the novel transverse fluid velocity component bears more information about the object shape. The classification of objects via hydrodynamic imaging thus benefits from 2D-sensing and can be scaled up to a supra biological scale of several meters. INDEX TERMS Hydrodynamic imaging, artificial lateral line, neural network, inverse problem, sensor array, feature selection.

show abstract

“…More directly, zero crossings, maxima, and minima of the velocity profile also directly produce estimates for the distance (y-coordinate) and the lateral position (x-coordinate) [43]. In [9,77,122], it is shown that an ALL comprising of 2D-sensitive sensors makes localization more robust. This extension compared to the 1D sensitivity of fish and most other ALL implementations, provides more and complementary spatial reference points [9,122], which may also be helpful for shape recognition.…”

Section: Velocity Profilesmentioning

confidence: 99%

Hydrodynamic Imaging with Artificial Intelligence: detecting submerged objects at a distance using a 2D-sensitive flow sensor array and neural networks

Wolf

View full text Add to dashboard Cite

Background 2 1.1.1 The biological lateral line 2 1.1.2 What is hydrodynamic imaging? 4 1.1.3 Artificial neural networks in practice 8 1.2 Research motivation 10 1.2.1 Scalability 10 1.2.2 Neural-network based signal processing 11 1.2.3 2D-sensitive sensing 11 1.3 Thesis outline 11 Graphical abstract 13 i sensing 2 design and validation of an all-optical 2d flow sensor 17 2.1 Introduction 18 2.2 Sensor and fluid model 20 2.2.1 Sensor design 20 2.2.2 FBG sensing 20 2.2.3 Isotropic sensor mechanics 21 2.2.4 Dynamic properties 24 2.2.5 Design optimization 25 2.3 Methods 26 2.4 Results 29 2.4.1 Linearity 29 2.4.2 Dynamic response 29 2.5 Discussion 31 2.5.1 Sensitivity and dynamic range 31 2.5.2 Two-dimensional fluid flow sensing 32 2.6 Conclusion 34 ii simulation studies 3 simulated 1d-sensitive localization 37 3.1 Introduction 38 3.2 Methods 40 v vi contents 3.2.1 Data generation 40 3.2.2 Neural network algorithms 44 3.3 Results 48 3.3.1 Parameter settings 48 3.3.2 Overall performance on high signal to noise input 49 3.3.3 MLP overfitting 50 3.3.4 Noise robustness 50 3.4 Discussion 53 3.4.1 Practical implementation neural networks 53 3.4.2 Overall performance on x and y coordinates 55 3.4.3 Noise robustness 55 3.5 Conclusion 56 3.5.1 Practical implementation neural networks 56 3.5.2 Further research on single source localization 56 4 simulated 1d-sensitive multi-source localization 59 4.1 Introduction 60 4.2 Background 61 4.2.1 Source location encoding by the lateral line organ 61 4.2.2 Object localization using artificial lateral lines 63 4.2.3 Convolutional neural networks 64 4.3 Methods 66 4.3.1 Location encoding in 3D 66 4.3.2 Data synthesis 69 4.3.3 CNN implementation 70 4.3.4 CNN optimization 73 4.3.5 Location decoding in 3D 75 4.4 Results 77 4.4.1 Probability grid reconstruction 77 4.4.2 3D position reconstruction 79 4.5 Discussion 79 4.5.1 The processing pipeline 80 4.5.2 CNN design variations 80 4.5.3 Iterative source detection 82 4.5.4 Detection limits and future research 83 4.6 Conclusion 84 Supplementary information 84 5 simulated 2d-sensitive localization 87 5.1 Introduction 88 5.1.1 State of the art 88 5.1.2 Aims of this study 89 contents vii 5.

show abstract

Training bioinspired sensors to classify flows

Cited by 27 publications

References 53 publications

Using machine learning to detect the turbulent region in flow past a circular cylinder

Using machine learning to detect the turbulent region in flow past a circular cylinder

Shape Classification Using Hydrodynamic Detection via a Sparse Large-Scale 2D-Sensitive Artificial Lateral Line

Hydrodynamic Imaging with Artificial Intelligence: detecting submerged objects at a distance using a 2D-sensitive flow sensor array and neural networks

Contact Info

Product

Resources

About