Deep learning velocity signals allows to quantify turbulence intensity
Abstract
Turbulence, the ubiquitous and chaotic state of fluid motions, is characterized by strong and statistically nontrivial fluctuations of the velocity field, over a wide range of length and timescales, and it can be quantitatively described only in terms of statistical averages. Strong nonstationarities hinder the possibility to achieve statistical convergence, making it impossible to define the turbulence intensity and, in particular, its basic dimensionless estimator, the Reynolds number.
Here we show that by employing Deep Neural Networks (DNN) we can accurately estimate the Reynolds number within accuracy, from a statistical sample as small as two largescale eddyturnover times. In contrast, physicsbased statistical estimators are limited by the rate of convergence of the central limit theorem, and provide, for the same statistical sample, an error at least times larger. Our findings open up new perspectives in the possibility to quantitatively define and, therefore, study highly nonstationary turbulent flows as ordinarily found in nature as well as in industrial processes.
Turbulence is characterized by complex statistics of velocity fluctuations correlated over a wide range of temporal and spatialscales. These range from the integral scale, , characteristic of the energy injection (with correlation time ), to the dissipative scale, , characteristic of the energy dissipation due to viscosity (with correlation time ). The intensity of turbulence directly correlates with the width of this range of scales, or , commonly dubbed inertial range.
In statistically stationary, homogeneous and isotropic turbulence (HIT), the width of the inertial range, is well known to correlate with the Reynolds number, , defined as , where is the characteristic velocity fluctuation at the integral scale, and is the kinematic viscosity. Therefore the value of the Reynolds number is customarily used to quantify turbulence intensity. While this value remains welldefined in laboratory experiments, performed under stationary conditions, and for fixed flow configurations (), its quantification results impossible when we consider turbulence in open environments (as in many outstanding geophysical situations) or in nonstationary situations (such as turbulent/nonturbulent interfaces). This observation is linked to the question: can we estimate turbulence intensity from fluctuating velocity signals of arbitrary (short) length? For statistically stationary conditions, this is indeed possible provided enough statistical samples are available and by using appropriate physicsbased statistical averages of the fluctuating velocity field. For nonstationary turbulent flows (i.e. changing on timescales comparable with the largescale correlation times) the question itself appears meaningless. In these conditions, the intertwined complexity of a slow largescale dynamics and of fast, but highly intermittent, smallscale fluctuations, makes impossible to estimate reliably the width of the inertial range.
In this paper we demonstrate, using a proof of concept, that our fundamental question can be answered by a suitable use of machine learning. We propose a machine learning Deep Neural Network (DNN) model capable of estimating turbulence intensity within accuracy from short velocity signals (duration : approximately two largescale eddy turnover times, i.e. , where ). We remark that analyzing the same data via standard statistical observables of turbulence leads to quantitatively meaningless results (predictions between and times the true value).
We train the DNN model to predict turbulence intensity using (short) Lagrangian velocity signals obtained from HIT. As Lagrangian velocities are one of the most intermittent features of turbulence, we are choosing the most difficult case for our proof of concept. The Lagrangian velocity signals, , that we employ are obtained as the superposition of different strongly chaotic time signals, , derived from a shell model of turbulence [biferale2003shell, l1998improved] (see Methods). The velocity signals, , are known to match the statistical properties of the velocity experienced by a passive Lagrangian particle in HIT [boffetta2002lagrangian]. The shell model describes the nonlinear energy transfer among different spatial scales, , where (with being the wave number associated with the integral scale, and defining the ratio between successive shells). The nonlinear energy transfer is characterized by sudden bursts of activity (typically referred to as “instantons”) [biferale2003shell, daumont2000instanton], where anomalous fluctuations are spread from large to smallscales. The complex spacetime patterns and localized correlations in , given by these intermittent bursts, make a 1dimensional Convolutional Neural Network a wellsuited choice for our neural network model (see Methods and SI for details).
We train the DNN using a collection of datasets corresponding to different viscosity values for our Lagrangian turbulent signal. Each dataset includes a large number of velocity signals (few thousands) sampled over time instants (see examples in Figure 1(a)). We decided to employ an external forcing to maintain the rootmeansquare energy fluctuations of the signals, and thus , statistically stationary. As a result, the viscosity fully determined the turbulence intensity, and therefore the Reynolds number. Decreasing the viscosity increases the highfrequency content of the velocity signals (as , cf. timeincrements in Figure 1(b)) by reducing the dissipative time and lengthscales. The resulting wider inertial range reflects the higher turbulence intensity. We train the network in a supervised way to infer the viscosity from the velocity signals; the collection of datasets covered uniformly the viscosity interval in equispaced levels.
We assess the predictive performance of the DNN by considering unseen signals generated by the shell model. Results are reported in Figure 2 where two different test sets are used: 1) “validation set”: including statistically independent realizations of the velocity signals than the training phase, yet having the same viscosity values; 2) “test set”: signals having different viscosity values than considered during training, yet within the same viscosity range. The results demonstrate that the network is capable of very accurate predictions for the viscosity over the full range considered, also for viscosity values different from those employed in training. By aggregating the viscosity estimates over a large number of statistically independent shell model signals with fixed viscosity, we can define an average estimate as well as a rootmeansquared error, see Figure 2(a).
To further reflect on this remarkable result, we turn to a physical argument for estimating the viscosity. For the secondorder Lagrangian structure functions , where , in steady HIT the following estimate holds in the small limit:
(1) 
where is a constant of order , , and is the Reynoldsindependent rate of energy dissipation [arneodo2008universal]. The (dissipative) timescale is defined by the relation which, using Eq. (1), gives . Assuming and known, or, alternatively estimating on a signalbysignal basis, to amend for largescale oscillations within the observation window, we evaluate by using a reference case of known viscosity, say . Figure 2(b,c,d) compare the pdfs of the logarithmic ratio between the estimated and true viscosity, for estimates by the DNN and based on Eq. (1). Predictions in case of Eq. (1) spread over a range , whereas this range is just of order in case of the DNN. Besides, in Figure 2(b), we observe that evaluating on a signalbysignal basis reduces, yet minimally, the variance of viscosity predictions based on Eq. (1).
The high variance and heavy tails of the pdf of viscosity estimates produced by Eq. (1) follow from the very limited statistical sampling ( points), which is severely affected by largescale oscillations and smallscale intermittent fluctuations. Because of these, statistical convergence and, therefore, a stable value for the LHS of Eq. (1), are attained only after very long observation times.
Our DNN model can be tested on real Lagrangian velocity signals, , obtained by the numerical integration of the Lagrangian dynamics of a tracer particle in HIT, see Figure 3. The underlying Eulerian velocity field is obtained from Direct Numerical Simulation [bec2010intermittency] of the NavierStokes equation at (see Methods). Remarkably, the DNN, although trained on shell model data, is able to estimate with extremely high accuracy the viscosity even in case of real Lagrangian data (note that Lagrangian velocity signals from DNS has been exhaustively validated against experimental data in the past [toschi2009lagrangian]). This points to the fact that the DNN relies on spacetime features that are equally present in the shell model as well as in the real Lagrangian signals.
What is the best result that can be achieved according to the current understanding of the physics of turbulence? Both by direct estimation of the Reynolds number or by viscous scale fitting, i.e. using Eq. (1), the statistical accuracy is limited by the fluctuations of the largescale velocity. Therefore, the statistical error is limited by the number of largescale eddy turnover times. As shown in Figure 2(b,c,d), a traditional statistical physics approach produces estimates for the viscosity spread over four orders of magnitude, while the DNN is capable of delivering accurate predictions, scattering within a range.
This points at two major results: first, the DNN, at least within the range of the training signals, must be able to identify spacetime structures that strongly correlate with turbulence intensity and which are rather insensitive to the strong fluctuations of the instantaneous value of the largescale velocity (cf. SI for a discussion). This finding unlocks the possibility of defining, practically instantaneously, turbulence intensities, Reynolds numbers, or connected statistical quantities for complex flows and fluids. Estimating locally, in space and in time, the turbulence intensity at laminarturbulent interfaces or from atmospheric anemometric readings can now be possible. The quantitative definition of an effective viscosity within boundary layers or complex fluids, such as emulsions or viscoelastic flows, may similarly be pursued. Finally, being able to extract the spacetime correlations identified by the DNN may give us novel and fundamental insights in turbulence physics and the complex skeleton of the fluctuating turbulent energy cascades.
Acknowledgments
The authors acknowledge the help of Pinaki Kumar for the development of the vectorized GPU code.
References
Methods
Generating the database of turbulent velocity signals
We employ the SABRA [l1998improved] shell model of turbulence to generate Lagrangian velocity signals corresponding to different turbulence levels (Reynolds numbers). Shell models evolve in time, , the complex amplitude of velocity fluctuations, , at logarithmically spaced wavelengths, ().
The amplitudes evolve according to the following equation:
(2) 
where represents the viscosity, is the forcing, and the real coefficients regulate the energy exchange between neighboring shells. We consider the following constraints: , which guarantees conservation of energy , for an unforced and inviscid system (, , respectively); , which gives to the second (inviscid/unforced) quadratic invariant of the system, , the dimensions of an helicity; to fix the third parameter we opt for the common choice . We truncate Eq. (2) to a finite number of shells which ensures a full resolution of the dissipative scales in combination with our forcing and viscosity range. We simulate the system in Eq. (2) via a RungeKutta scheme with viscosity explicitly integrated [bohr2005dynamical] (the integration step, , is fixed for all simulations, to be about three orders of magnitude smaller than the dissipative timescale for the lowest viscosity case).
We inject energy through a largescale forcing acting on the first two shells [l1998improved]. The forcing dynamics is given by an OrnsteinUhlenbeck process with a timescale matching the eddy turnover of the forced shells (, ). Additionally, we set the ratio between the standard deviation () of the two forcing signals. This ensures a helicityfree energy flux in the system [l1998improved]. See the SI for further information on the signals and values of the constants.
We generate the signals in a vectorized fashion on an nVidia V100 card. We integrate simultaneously instances of the system in Eq. (2) in a vectorized manner (i.e. system description by complex variables), and dump the state times after skipping the first samples.
Lagrangian velocity signals from Direct Numerical Simulations
The true Lagrangian velocity signals are obtained from the numerical integration of Lagrangian tracers dynamics evolved on top of a Direct Numerical Simulation of HIT turbulence. The Eulerian flowfield is evolved via a fully dealiased algorithm with secondorder AdamsBashfort timestepping with viscosity explicitly integrated. The Lagrangian dynamics is obtained via a trilinear interpolation of the Eulerian velocity field coupled with secondorder AdamsBashfort integration in time. The Eulerian simulation has a resolution of grid points, a viscosity of , a timestep , this corresponded to a , dissipative scale and . The Lagrangian trajectories employed are available at the 4TU.Centre for Research Data [LagrDNSDATA].
Deep Neural Network (DNN)
We employ a onedimensional Convolutional Neural Network (CNN) architecturally inspired by the VGG model [simonyan2014very]. Developing a neural network model poses the major challenge of selecting a large number of hyperparameters. This particular architecture deals with this issue by fixing the size of the filters and employs stacks of convolutional layers to achieve complex detectors. For our model we opted for convolutional filters of size , which is comparable or smaller than the dissipative timescale of the turbulent signals. The network includes four blocks, each formed by three convolutional layers (including filters each), a max pooling layer (window: 2) and a dropout layer, that capture all the spatial features of the signal (cf. DNN architecture in SI). These layers are followed by a fullyconnected layer with neurons and ReLu activation that collects all the spatial features into a dense representation. The final layer provides a linear map from the dense representation to the estimated viscosity. A complete sketch of the network is in the SI.
DNN Training
We train the neural network in a supervised fashion and with training loss to output a continuous value in the interval , which is linearly mapped to . The training set is composed of turbulent velocity signals (timesampled over points) uniformly distributed among viscosity levels (trainingvalidation ratio: 75%25%). See Table in SI for further information.
Supplementary Information (SI)
Width of the inertial range
In presence of limited statistics as in the case of relatively short signals, the estimation of the width of the inertial range or, similarly, the estimation of the viscosity, is enslaved to large scale energy fluctuations. On a time scale comparable to the large scale fluctuations, local increments or decrements of the system energy yield almost instantaneous widenings or shortenings of the inertial range. This effect can be naturally interpreted in terms of viscosity, where local energy increments play the same effect of a lower viscosity on the width of inertial range (see Figure S.1, where show this aspect for Eulerian structure functions).
In Figure S.2(a), we report Lagrangian structure functions for a set of training signals with fixed viscosity values. The limited statistics yield high fluctuations among the structure function, due to a combination of largescale energy fluctuations and small scale intermittency. In Figure S.2(b), we amend largescale fluctuations by normalizing by the signal energy, i.e. we report .
Data generation, training, testing and neural network parameters
We include in Table 1 the parameters considered in the shell model simulations by which the training, validation and test datasets have been created. In Figure S.3 we complement Figure 1 by including, for the same three viscosity levels, further features of the considered signals. These are: (a) second order Eulerian structure functions, (where ) showing that changing the viscosity only affects the extension of the inertial range; (b) relevant time scales (computed by inertial scaling) associated with the dynamics of the different shells; (c) signals energy as a function of time. In Figure S.4, we report the diagram of the neural network. Relevant structural parameters (e.g. size of the convolutional filters) are reported in the figure caption.
Parameter  Value  

Number of shells  
Wave number integral scale  
2  Intershell distance  
Noise intensity forcing on shell  
” on shell  
integration step  
sampling time DNN  
length window DNN 
Parameter  Training  Testing 

increment  
levels  
set size  192.000  6.600 
training:validation ratio  75%:25%  N/A 

Features observed by the DNN
During training, the DNN develops feature detectors. As discussed in the main text, we expect these detectors to select features that, at the same time, strongly correlate with the turbulence intensity and that are insensitive to large scale oscillations. As generally expected in deep learning, detectors are likely specific to the parameter range and statistical properties of the signals contained in the training set.
In this section, to understand the characteristics of the signal that our model relies on, we develop an ablation study by systematically altering the content of randomly selected testing signals. The modifications considered involve the suppression of frequency components, or the random shuffling of the time structure. This enables us to identify features mostly ignored by the DNN and, conversely, restrict the set of characteristics of the signals relevant for the DNN.
In Figure S.5 we consider testing signals that have been altered through a highpass (a) or a bandpass filter (b). In the case of Lagrangian signals, filtering operations are easily performed by restricting the summation in Eq. (2) to a subset of the shell signals. We select one testing signal per viscosity level, we ablate its spectral structure and we plot the DNN prediction. We notice that the neural network is almost insensitive to the large scale dynamics, as the estimates after the highpass filter remain unaltered if the largescale shells are removed. We notice, in particular, that any selection of a band of shells that includes the last part of the inertial range yield almost errorfree predictions.
Similarly, we can alter the time structure of the signals by partitioning them in disjoint contiguous blocks of length , and then by randomly mixing these blocks. In Figure S.6 we report the predictions for different block extensions. As the block extension remains in the same order of the integral scale, the prediction remain mostly unaltered, to then degrade as the block size become comparable to the dissipative timescale. This shows how the training develop feature extractors targeting fine scales and correlations existing around the dissipative end of the inertial range.