A Novel Deep Learning Model for the Flow Field Reconstruction of An Oscillating Airfoil

In this paper, a novel model is presented for reconstructing unsteady periodic fields of velocity vector and pressure scalar over an oscillating foil. This data-driven method based on convolutional neural network can be utilized to accomplish two objections: fields reconstruction from limited measurements and transient aerodynamic characteristics prediction. The verification results of an oscillating foil under low Reynolds number show that this method can accurately reconstruct all the fields only by limited pressure information at probes on the foil surface. The evaluation on aerodynamic characteristics prediction illustrates that our model outperforms four classical machine learning methods. Meanwhile, a well-trained CNN model can almost achieve real-time flow field prediction by leveraging the GPU acceleration. Finally, the exploration of the robustness for the CNN model is conducted on several aspects, including training size, probe layouts, probe numbers and measurement noises.

Adaptive POD surrogate model method for centrifugal pump impeller flow field reconstruction based on clustering algorithm

To reduce the calculation cost and improve the accuracy of flow field prediction, an adaptive proper orthogonal decomposition (APOD) surrogate model based on K-means clustering algorithm was proposed to reconstruct the flow field of impeller. The experiment samples were designed by introducing the perturbation of the blade control parameters such as blade wrap angle and blade angle of outlet. K-means clustering algorithm was used to classify the sample blade shapes, and find out the cluster of the objective blade. The snapshot set, which consisted of the blade shape and the flow field data of impeller, can be described as a linear combination of orthogonal basis by POD method. The radial basis function (RBF) was used to fit the orthogonal basis coefficients of the objective blade, and then the flow field of objective impeller was reconstructed. The traditional fixed sample POD (FPOD) method and the proposed APOD method were used to reconstruct the flow field in impeller, respectively, and the prediction results of the two methods were compared and analyzed. The results show that the proposed APOD method could quickly and accurately reconstruct the objective flow field. The flow field prediction accuracy of the APOD method is significantly higher than the FPOD method, and the calculation time for the flow field prediction is less than 1/360 of the CFD.

How Euglena gracilis swims: flow field reconstruction and analysis

Euglena gracilis is a unicellular organism that swims by beating a single anterior flagellum. We study the nonplanar waveforms spanned by the flagellum during a swimming stroke, and the three-dimensional flows that they generate in the surrounding fluid.Starting from a small set of time-indexed images obtained by optical microscopy on a swimming Euglena cell, we construct a numerical interpolation of the stroke. We define an optimal interpolation (which we call synthetic stroke) by minimizing the discrepancy between experimentally measured velocities (of the swimmer) and those computed by solving numerically the equations of motion of the swimmer driven by the trial interpolated stroke. The good match we obtain between experimentally measured and numerically computed trajectories provides a first validation of our synthetic stroke.We further validate the procedure by studying the flow velocities induced in the surrounding fluid. We compare the experimentally measured flow fields with the corresponding quantities computed by solving numerically the Stokes equations for the fluid flow, in which the forcing is provided by the synthetic stroke, and find good matching.Finally, we use the synthetic stroke to derive a coarse-grained model of the flow field resolved in terms of a few dominant singularities. The far field is well approximated by a time-varying Stresslet, and we show that the average behavior of Euglena during one stroke is that of an off-axis puller. The reconstruction of the flow field closer to the swimmer body requires a more complex system of singularities. A system of two Stokeslets and one Rotlet, that can be loosely associated with the force exerted by the flagellum, the drag of the body, and a torque to guarantee rotational equilibrium, provides a good approximation.

Flow field reconstruction method based on array neural network

ABSTRACTMulti-dimensional aerodynamic database technology is widely used, but its model often has the curse of dimensionality. In order to solve this problem, we need projection to reduce the dimension. In addition, due to the lack of traditional method, we have improved the traditional flow field reconstruction method based on artificial neural networks, and we proposed an array neural network method.In this paper, a set of flow field data for the target problem of the fixed Mach number is obtained by the existing CFD method. Then we arrange all the sampled flow field data into a matrix and use proper orthogonal decomposition (POD) to reduce the dimension, whose size is determined by the first few modals of energy. Therefore, significantly reduced data are obtained. Then we use an arrayed neural network to map the flow field data of simplified target problem and the flow field characteristics. Finally, the unknown flow field data can be effectively predicted through the flow field characteristic and the trained array neural network.At the end of this paper, the effectiveness of the method is verified by airfoil flow fields. The calculation results show that the array neural network can reconstruct the flow field of the target problem more accurately than the traditional method, and its convergence speed is significantly faster. In addition, for the case of high angle flow field, the array neural network also performs well. There are no obvious jumps, and huge errors are found in results. In general, the proposed method is better than the traditional method.