Zero-shot action recognition (ZSAR) aims to recognize unseen action categories in the test set without corresponding training examples. Most existing zero-shot methods follow the feature generation framework to transfer knowledge from seen action categories to model the feature distribution of unseen categories. However, due to the complexity and diversity of actions, it remains challenging to generate unseen feature distribution, especially for the cross-dataset scenario when there is potentially larger domain shift. This paper proposes a De confounding Ca usa l GAN (DeCalGAN) for generating unseen action video features with the following technical contributions: 1) Our model unifies compositional ZSAR with traditional visual-semantic models to incorporate local object information with global semantic information for feature generation. 2) A GAN-based architecture is proposed for causal inference and unseen distribution discovery. 3) A deconfounding module is proposed to refine representations of local object and global semantic information confounder in the training data. Action descriptions and random object feature after causal inference are then used to discover unseen distributions of novel actions in different datasets. Our extensive experiments on C ross- D ataset Z ero- S hot A ction R ecognition (CD-ZSAR) demonstrate substantial improvement over the UCF101 and HMDB51 standard benchmarks for this problem.

The understanding and controlling of the restrained shrinkage is critical for the application of ultra high performance concrete (UHPC). This study made an analysis of restrained shrinkage mechanism of four kinds of UHPCs based on the direct tensile test accompanied with acoustic emission (AE) source location, free shrinkage test and restrained shrinkage test. The effects of UHPC tensile properties (strain softening or strain hardening) and high performance calcium sulphoaluminate (HCSA, a new kind of expansion agent) dosages (0%, 3% and 6% by mass of total binder) were investigated. The restrained shrinkage cracking mechanism of strain hardening UHPC and strain softening UHPC was analyzed based on AE analysis method under direct tensile loading. The results indicates that strain hardening UHPC shows multiple micro-defects to relax the restrained tensile stress marginally while strain softening UHPC shows several hairline cracks to relax the restrained tensile stress evidently. In ASTM C1581-04 ring test, the restrained shrinkage of strain hardening UHPC with HCSA dosage of 0%, 3% and 6% at 80d is 141 με, 96 με and 16 με, respectively. The HCSA expansion agent can effectively reduce the restrained shrinkage of UHPC and the influence on the structure.

Water transport is vital for the durability of ultra-high performance concrete (UHPC) in engineering, but its absorption behavior requires further comprehension. This study investigates the impact of silica fume (SF) and metakaolin (MK) on water absorption in UHPC matrix with a high volume of limestone powder (LS) under two curing temperatures, and the variation in water transport with pore size obtained by low field nuclear magnetic resonance (LF-NMR). Relations between cumulative water absorption with other properties were discussed, and the pore size distribution (PSD) measured by Mercury intrusion porosimetry (MIP) was compared with that determined by LF-NMR. Results showed that MK outperformed SF in reducing water absorption in UHPC matrix, containing 30% LS under steam curing due to the synergistic effect between MK and LS. The incorporation of LS greatly affected the water absorption process of UHPC matrix. In samples without LS, capillary and gel pores absorbed water rapidly within the first 6 h and slowly from 6 h to 48 h simultaneously. However, in samples with 30% LS, gel pore water decreased during water absorption process due to the coarsening of gel pores. MK was able to suppress gel pore deterioration caused by the addition of a large amount of LS. Compared with PSD measured by MIP, NMR performed better in detecting micropores (<10 nm).

The tensile stress–strain response is considered to be the most important and fundamental mechanical property of ultra-high-performance fiber-reinforced concrete (UHPFRC). Nevertheless, it is still a challenging matter for researchers to determine the tensile properties of UHPFRC. As a simpler alternative to the direct tensile test, bending tests are widely performed to characterize the tensile behavior of UHPFRC, but require further consideration and a sophisticated inverse analysis procedure. In order to efficiently predict the tensile properties of UHPFRC, a nonlinear inverse method based on notched three-point bending tests (3PBT) was proposed in this paper. A total of fifteen UHPFRC beams were fabricated and tested to evaluate the sensitivity of the predicted tensile behavior to variations in fiber volume fraction. A segmented stress–strain model was used, which is capable of describing the various tensile properties of UHPFRC, including strain softening and strain hardening. A more approximate formulation was adopted to simulate the load–deflection response of UHPFRC beam specimens. The closed-form analytical solutions were validated by tensile test results and existing methods in literature. Finally, parametric studies were also conducted to investigate the robustness of the proposed method. The load–deflection responses obtained from notched 3PBT could be easily converted into tensile properties with this inverse method.