Search

In order to improve the availability of waste rubber tires and bearing capacity of sand foundation, a new composite foundation transforms the tensile strength of a discarded rubber tire into a hoop-tightening force on granular sand soil, i.e. hoop capacity, is introduced and investigated. Considering the characteristics of the circular enclosed structure of the tire, the tensile strength of the tire strips was obtained by carrying out a tension test of tire strips, and the relationships between earth pressure, settlement and vertical load were discovered by the results of loading tests under the conditions of confined and unconfined. The distribution of circumferential stress at each position of the tire cross section was monitored to evaluate the contribution of each part of the tire to the integral hoop capacity. Based on the theory of annular elasticity, Airy stress-state function is introduced to determine the solutions of stress, strain and displacement in the sand and tire regions based on the stress boundary condition and displacement continuity condition. The results show that the theory prediction curve has a high precision approaching the measured data, which indicates the rationality of the proposed theoretical model in predicting the circumferential stress of this type of waste rubber tire composite foundations. [ABSTRACT FROM AUTHOR]

Biomedical named entity recognition (Bio-NER) is the prerequisite for mining knowledge from biomedical texts. The state-of-the-art models for Bio-NER are mostly based on bidirectional long short-term memory (BiLSTM) and bidirectional encoder representations from transformers (BERT) models. However, both BiLSTM and BERT models are extremely computationally intensive. To this end, this paper proposes a temporal convolutional network (TCN) with a conditional random field (TCN-CRF) layer for Bio-NER. The model uses TCN to extract features, which are then decoded by the CRF to obtain the final result. We improve the original TCN model by fusing the features extracted by convolution kernel with different sizes to enhance the performance of Bio-NER. We compared our model with five deep learning models on the GENIA and CoNLL-2003 datasets. The experimental results show that our model can achieve comparative performance with much less training time. The implemented code has been made available to the research community.

The interaction between viral proteins and small molecule compounds is the basis of drug design. Therefore, it is a fundamental challenge to identify viral proteins according to their amino acid sequences in the field of biopharmaceuticals. The traditional prediction methods su er from the data imbalance problem and take too long computation time. To this end, this paper proposes a deep learning framework for virus protein identifying. In the framework, we employ Temporal Convolutional Network(TCN) instead of Recurrent Neural Network(RNN) for feature extraction to improve computation e ciency. We also customize the cost-sensitive loss function of TCN and introduce the misclassification cost of training samples into the weight update of Gradient Boosting Decision Tree(GBDT) to address data imbalance problem. Experiment results show that our framework not only outperforms traditional data imbalance methods but also greatly reduces the computation time with slight performance enhancement.