Publications

This note presents the main process of optimization design of foam core/carbon fiber composite sandwich which primarily designed for UAV wing beams. During the actual application, the original design provided excessive structural strength and it has certain capacity to be optimized. So the weight of structure can be reduced under the premise of meet the strength requirement. In order to characterize fully the complex mechanical behavior of such a highly heterogeneous material and find the ultimate strength of this structure, MSC.Patran/Nastran has be applied on analysis of this composite sandwich structure. Base on the result of the numerical simulations, the best combination of composite laminated and the material layer thickness have be determined, and the beams structure of the lightweight has be designed ultimately.

Attitude Determination (AD) of a Low Earth Orbit (LEO) small satellite using the code and phase measurements of Global Navigation Satellite Systems (GNSS) is explored in this research. The proliferation of satellite based navigation system and the burgeoning application area of small satellites in the field of Earth observation has motivated this research paradigm. This contribution focuses on the design, simulation and comparison of multi-antenna, stand-alone, commercial-off-the-shelf (COTS), GPS-based attitude determination using the code and phase measurements of GPS signals. Master antenna position is obtained by Single Point positioning method using the code observations and ephemeris while differential positioning is applied to find the slave antenna positions. The problem of Ambiguity Resolution (AR) that is crucial while estimating attitude using phase measurements is done by Least-Squares AMBiguity Decorrelation Adjustment (LAMBDA) method. Finally, the attitude is determined using the direct and the least square method. It has been found that phase measurement method is more precise than the code based method and an accuracy of less than one degree in attitude parameters can be obtained using phase measurements.

The present paper focuses on the analysis and processing of the observational data by using the high-precision processing software gamit/globk,and the contrast of the effects of broadcast ephemeris and precise ephemeris on the LC phase residuals of GPS monitoring stations.Under the same conditions,the orbital errors by using the broadcast ephemeris tend to be 1.5 to 27.2 mm greater than those by using the precise ephemeris.These differences have something to do with the number and length of the short baselines in the stations.Through the data processing by using the dual-frequency differential analysis,for the baselines with the length of 300 km or less,the baseline errors generally range between 5 and 8 mm as using the broadcast ephemeris while the baseline errors vary from 3.2 to 6.8 mm as using the precise ephemeris;the differences of the baseline errors range between 1.2 and 1.8 mm.It can be seen from the algorithm case study in the text that both the broadcast ephemeris and precise ephemeris are well solvable when the baselines lengths come up to 300 km or less,which have increased by a big margin than the baseline length of 100 km proposed by the previous research workers.The authors in this study owe all the progress to the application of the high-precision software gamit/globk and initial coordinates,suitable constracts,effective use of the observational data and processing models.

The reconstruction of Nearfield Acoustic Holography(NAH) is a linear,ill-posed inverse problem,in which a regularization procedure must be used.We compared three Tikhonov regularization based parameter selection methods, Morozov Discrepancy Principle method(MDP),Generalized Cross Validation method(GCV),and L-curve method,at various hologram distances,sound source frequencies,and Signal-to-Noise Ratios(SNR).The results show that they are all not robust at large hologram distances and low noise conditions.Based on an equivalent-noise-variance method,we established an improved method for estimating the noise variance in the MDP method,which is relative robust among the three,and made it work at relatively large hologram distances and low noise conditions.Numerical simulations show that the improved MDP method is robust in determining proper regularization parameters over a large SNR range(6 dB) and at a relatively large hologram distance(～10 cm).Moreover,the improved MDP method has a valid reconstruction aperture that is equal to the hologram aperture because it is not necessary to smooth the hologram pressure.

Strictly enforcing orthonormality constraints on parameter matrices has been shown advantageous in deep learning. This amounts to Riemannian optimization on the Stiefel manifold, which, however, is computationally expensive. To address this challenge, we present two main contributions: (1) A new efficient retraction map based on an iterative Cayley transform for optimization updates, and (2) An implicit vector transport mechanism based on the combination of a projection of the momentum and the Cayley transform on the Stiefel manifold. We specify two new optimization algorithms: Cayley SGD with momentum, and Cayley ADAM on the Stiefel manifold. Convergence of Cayley SGD is theoretically analyzed. Our experiments for CNN training demonstrate that both algorithms: (a) Use less running time per iteration relative to existing approaches that enforce orthonormality of CNN parameters; and (b) Achieve faster convergence rates than the baseline SGD and ADAM algorithms without compromising the performance of the CNN. Cayley SGD and Cayley ADAM are also shown to reduce the training time for optimizing the unitary transition matrices in RNNs.

Aimed at the gear fault diagnosis, a diagnosis system which based on the wavelet for picking up character and BP neural network are proposed, the energy distributing of each frequency segment which is decomposed by wavelet are treated as the eigenvector and input the NN. The testing result indicates that this method can accurately diagnose the fault of gear and has extensive application foreground.

Abstract After fracture treatment in unconventional reservoirs, the in-situ stress and fluid pressure are greatly changed in the reservoir because of the generation of fracture networks. In order to get high production, efforts are made to get close fracture spacing and long fracture length in-situ field, which in turn make fracture distribution become complicated as the range of fractures size and density is widespread. In this work, the finite element method is used to analysis flowback around hydraulic fracture among complex fractures networks, which consider the coupled effects of flow and geomechanics. The reservoir is assumed to be a 3-D poroelastic medium. According to the fracture sizes, the fracture is divided into three types. These small natural fractures are treated as SRV regions, hydraulic fractures, natural fracture in middle and large sizes are explicitly represented using LGR. Finite element method simulates fracture deformation and the two-phase fluid flow in the reservoir during flowback stage. The physical properties are altered by the coupled flow and geomechanics in the reservoir. The fluid pressure, stress and flowback production over time around these fractures are recorded. The results show that during the flowback period, the production experience a sharp decrease. The porosity and permeability in the reservoir are greatly reduced because of the coupled effects. These explicit natural fractures influence the hydraulic fractures. As the hydraulic fracture spacing reduced, the stress shadow effects become more serious and the flowback production decreases. This work helps understand the flowback analysis with coupled geomechanics and flow effects in the complex fracture networks in the unconventional reservoirs and physical properties effects in different reservoir conditions.

An accurate trajectory prediction is crucial for safe and efficient autonomous driving in complex traffic environments. In recent years, artificial intelligence has shown strong capabilities in improving prediction accuracy. However, its characteristics of inexplicability and uncertainty make it challenging to determine the traffic environmental effect on prediction explicitly, posing significant challenges to safety-critical decision-making. To address these challenges, this study proposes a trajectory prediction framework with the epistemic uncertainty estimation ability that outputs high uncertainty when confronting unforeseeable or unknown scenarios. The proposed framework is used to analyze the environmental effect on the prediction algorithm performance. In the analysis, the traffic environment is considered in terms of scenario features and shifts, respectively, where features are divided into kinematic features of a target agent, features of its surrounding traffic participants, and other features. In addition, feature correlation and importance analyses are performed to study the above features' influence on the prediction error and epistemic uncertainty. Further, a cross-dataset case study is conducted using multiple intersection datasets to investigate the impact of unavoidable distributional shifts in the real world on trajectory prediction. The results indicate that the deep ensemble-based method has advantages in improving prediction robustness and estimating epistemic uncertainty. The consistent conclusions are obtained by the feature correlation and importance analyses, including the conclusion that kinematic features of the target agent have relatively strong effects on the prediction error and epistemic uncertainty. Furthermore, the prediction failure caused by distributional shifts and the potential of the deep ensemble-based method are analyzed.