Search

Understanding and predicting riverbed deformation during floods is important for the proper and sustainable design of headworks structures. Field-scale study of floods is challenging, and existing literature shows that mainly qualitative data (Photographic and Videographic) have been used in riverbed deformation studies. On the other hand, laboratory experiments allow studying floods in more detail and under a controlled environment. In this research, bed deformation of river channel during floods was investigated by conducting experiments in a curved flume. Parameters of the experiments were chosen with respect to gravel-cobbled River in Nepal. Four different floods with different magnitudes and durations were investigated in the experiments. Riverbed deformation was quantified by differencing pre and post-flood DEMs obtained using Structure from Motion (SfM) technology. The maximum scour during the floods was measured by scour chains made with light beads in a thread. The results showed that the magnitude and duration of floods are both important factors for riverbed deformation. The volume of channel bed erosion increased with the increase in flood magnitude. However, the total volume of sediment reworked was higher in the case of floods which are of lower magnitude but occur frequently in multiple cycles. The depth of maximum bed scour during the floods was observed much higher than the channel bed resulted at the end of the floods. The difference was more prominent along the bends where the secondary currents are stronger. The understanding could be used to better validate sediment transport models, and thus predict riverbed deformation more accurately for sustainable design of headworks structures.

Data augmentation is a valuable tool for the design of deep learning systems to overcome data limitations and stabilize the training process. Especially in the medical domain, where the collection of large-scale data sets is challenging and expensive due to limited access to patient data, relevant environments, as well as strict regulations, community-curated large-scale public datasets, pretrained models, and advanced data augmentation methods are the main factors for developing reliable systems to improve patient care. However, for the development of medical acoustic sensing systems, an emerging field of research, the community lacks large-scale publicly available data sets and pretrained models. To address the problem of limited data, we propose a conditional generative adversarial neural network-based augmentation method which is able to synthesize mel spectrograms from a learned data distribution of a source data set. In contrast to previously proposed fully convolutional models, the proposed model implements residual Squeeze and Excitation modules in the generator architecture. We show that our method outperforms all classical audio augmentation techniques and previously published generative methods in terms of generated sample quality and a performance improvement of 2.84% of Macro F1-Score for a classifier trained on the augmented data set, an enhancement of $1.14\%$ in relation to previous work. By analyzing the correlation of intermediate feature spaces, we show that the residual Squeeze and Excitation modules help the model to reduce redundancy in the latent features. Therefore, the proposed model advances the state-of-the-art in the augmentation of clinical audio data and improves the data bottleneck for the design of clinical acoustic sensing systems.