Generalizability of reservoir computing for flux-driven two-dimensional convection.

We explore the generalization properties of an echo state network applied as a reduced-order model to predict flux-driven two-dimensional turbulent convection. To this end, we consider a convection domain with constant height with a variable ratio of buoyancy fluxes at the top and bottom boundaries, which break the top-down symmetry in comparison to the standard Rayleigh-Bénard case, thus leading to highly asymmetric mean and fluctuation profiles across the layer. Our direct numerical simulation model describes a convective boundary layer in a simple way. The data are used to train and test a recurrent neural network in the form of an echo state network. The input of the echo state network is obtained in two different ways, either by a proper orthogonal decomposition or by a convolutional autoencoder. In both cases, the echo state network reproduces the turbulence dynamics and the statistical properties of the buoyancy flux, and is able to model unseen data records with different flux ratios.

Echo state network for two-dimensional turbulent moist Rayleigh-Bénard convection.

Recurrent neural networks are machine learning algorithms that are well suited to predict time series. Echo state networks are one specific implementation of such neural networks that can describe the evolution of dynamical systems by supervised machine learning without solving the underlying nonlinear mathematical equations. In this work, we apply an echo state network to approximate the evolution of two-dimensional moist Rayleigh-Bénard convection and the resulting low-order turbulence statistics. We conduct long-term direct numerical simulations to obtain training and test data for the algorithm. Both sets are preprocessed by a proper orthogonal decomposition (POD) using the snapshot method to reduce the amount of data. Training data comprise long time series of the first 150 most energetic POD coefficients. The reservoir is subsequently fed by these data and predicts future flow states. The predictions are thoroughly validated by original simulations. Our results show good agreement of the low-order statistics. This incorporates also derived statistical moments such as the cloud cover close to the top of the convection layer and the flux of liquid water across the domain. We conclude that our model is capable of learning complex dynamics which is introduced here by the tight interaction of turbulence with the nonlinear thermodynamics of phase changes between vapor and liquid water. Our work opens new ways for the dynamic parametrization of subgrid-scale transport in larger-scale circulation models.

Impairment of renal function using hyperoncotic colloids in a two hit model of shock: a prospective randomized study.

INTRODUCTION: One of the therapeutic essentials in severe sepsis and septic shock is an adequate fluid replacement to restore and maintain circulating plasma volume, improve organ perfusion and nutritive microcirculatory flow. The type of solution to be used as a fluid replacement remains under discussion. The aim of the study was to evaluate the effects of clinically used fluid replacement solutions on renal function and inflammatory response. METHODS: A total of 23 anesthetized and ventilated female German Landrace pigs were investigated over 19 hours using a two-hit model that combined hemorrhagic and septic shock. The septic shock was induced using an Escherichia coli laden clot placed into the abdominal cavity. Infusions of 6% hydroxyethylstarch 130/0.42 in acetate (6% HES 130), 4% gelatin in acetate (4% gelatin) and 10% hydroxyethylstarch 200/0.5 in saline (10% HES200) compared to Ringer's acetate (RAc) were used for fluid replacement to maintain a central venous pressure of 12 mmHg. Ringer's acetate was also used in the sham-treated group (SHAM). RESULTS: At study end the cardiac output (10% HES200 143±48 ml/kgBW; 6% HES130 171±47 ml/kgBW; RAc 137±32 ml/kgBW; 4% gelatin 160±42 ml/kgBW), as well as mean arterial pressure did not differ between groups. N-acetyl-beta-D-glucosamidase was significantly higher in the hydroxyethylstarch 200 (157±115 U/g creatinine; P<0.05) group compared to hydroxyethylstarch 130 (24±9 U/g creatinine), Ringer's acetate (2±3 U/g creatinine) and SHAM (21±15 U/g creatinine) at the study's end. Creatinine significantly increased by 87±84 percent of baseline in the 10% HES200 group compared to RAc and 6% HES130. We demonstrated in the histology of the kidneys a significant increase in osmotic-nephrosis like lesions for 4% gelatin compared to RAc, 6% HES130 and SHAM. Urine output was lowest in the 10% HES200 and 4% gelatin group, however not significantly.Interleukin(IL)-6 levels were significantly elevated in the 10% HES200 group (3,845±1,472 pg/ml) two hours after sepsis induction compared to all other groups (6% HES130 1,492±604 pg/ml; RAc 874±363 pg/ml; 4% gelatin 1,623±1,242 pg/ml). CONCLUSIONS: Despite similar maintenance of macrocirculation 6% hydroxyethylstarch 130/0.42 and Ringer's acetate significantly preserve renal function and attenuate tubular damage better than 10% hydroxyethylstarch 200/0.5 in saline.