SUMMARYPseudo-dynamic tests on a large-scale model of an existing six-pier bridge were performed at the ELSA laboratory using the substructuring technique. Two physical pier models were constructed and tested in the laboratory, while the deck, the abutments and the remaining four piers were numerically modeled on-line. These tests on a large-scale model of an existing bridge are the ÿrst to have been performed considering non-linear behavior for the modeled substructure. Asynchronous input motion, generated for the speciÿc bridge site, was used for the abutments and the pier bases. Three earthquake tests with increasing intensities were carried out, aimed at the assessment of the seismic vulnerability of a typical European motorway bridge designed prior to the modern generation of seismic codes. The experimental results conÿrm the poor seismic behavior of the bridge, evidenced by irregular distribution of damage, limited deformation capacity, tension shift e ects and undesirable failure locations.

No abstract

SUMMARYThe paper describes the distinctive features of the pseudo-dynamic test method as implemented at the ELSA reaction-wall facility. Both hardware and software aspects are considered. Particular attention is devoted to the digital control system and to a coupled numerical}experimental substructuring technique allowing realistic earthquake testing of very large structures. Mathematical and implementation details corresponding to this testing technique are given for both synchronous and asynchronous input motion. Selected test results illustrate the advantages of the presented features.

Automatically finding good and general remote sensing representations allows to perform transfer learning on a wide range of applications -improving the accuracy and reducing the required number of training samples. This paper investigates development of generic remote sensing representations, and explores which characteristics are important for a dataset to be a good source for representation learning. For this analysis, five diverse remote sensing datasets are selected and used for both, disjoint upstream representation learning and downstream model training and evaluation. A common evaluation protocol is used to establish baselines for these datasets that achieve state-of-the-art performance. As the results indicate, especially with a low number of available training samples a significant performance enhancement can be observed when including additionally in-domain data in comparison to training models from scratch or fine-tuning only on ImageNet (up to 11% and 40%, respectively, at 100 training samples). All datasets and pretrained representation models are published online.

The work presented in this paper is part of a research project aiming for the development of a performance-based approach for sustainable design, focusing on the efficient use of natural resources over the lifetime of buildings. The proposed approach requires the set of benchmarks to provide a consistent and transparent yardstick for the environmental performance of buildings and to strive towards an effective reduction in the use of resources and relative environmental impacts in the building sector. This paper focuses on the development of the framework for the quantification of the benchmarks. Additionally, a review of available benchmarks is provided, showing a huge diversity of values. One of the main factors contributing to such diversity is the lack of a reliable model for the quantification of the benchmarks. To overcome this problem, a consistent model for life-cycle assessment (LCA) is adopted, which is based on a standardized framework and enables comparability of results. Based on the proposed approach, a preliminary set of benchmarks for residential buildings is defined, leading to values in the range of 5-12 kg CO 2 /m 2 .yr and 68-186 MJ/m 2 .yr, for life-cycle global warming and total primary energy, respectively.