Search for: All records

Seismic resiliency includes the ability to protect the contents of mission-critical buildings from becoming damaged. The contents include telecommunication and other types of electronic equipment in mission-critical data centres. One technique to protect sensitive equipment in buildings is the use of floor isolation systems (FIS). Multi-directional shake table real-time hybrid simulation (RTHS) is utilized in this paper to validate the performance of full-scale rolling pendulum (RP) bearings, incorporating multi-scale (building– FIS–equipment) interactions. The analytical substructure for the RTHS included 3D nonlinear models of the building and isolated equipment, while the experimental substructure was comprised of the FIS. The RTHS test setup consisted of the FIS positioned on a shake table, where it is coupled to the analytical substructure and subjected to multi-directional deformations caused by the building’s floor accelerations and equipment motion from an earthquake. Parametric studies were performed to assess the influence of different building lateral load systems on the performance of the FISs. The lateral load resisting systems included buildings with steel moment resisting frame (SMRF) systems and with buckling restrained braced frame (BRBF) systems. Each building type was subjected to multi-directional ground motions of different sources and hazard levels. Details of the experimental test setup, RTHS test protocol and main preliminary results on the multi-directional testing of an RP-based FIS are described. Challenges in conducting the multi-axial RTHS, including the nonlinear kinematics transformation, adaptive compensation for the actuator-table dynamics, along with the approaches used to overcome them are presented. The acceleration and deformation response of the isolated equipment is assessed to demonstrate the effectiveness of the FIS in mitigating the effects of multi-directional seismic loading on isolated equipment in mission-critical buildings. 

The process of training deep learning models produces a huge amount of meta-data, including but not limited to losses, hidden feature embeddings, and gradients. Model diagnosis tools have been developed to analyze losses and feature embeddings with the aim to improve the performance of these models. However, gradients, despite carrying rich information that is potentially relevant for model interpretation and data debugging, have yet to be fully explored due to their size and complexity. Each single gradient has a size as large as the number of parameters of the neural net - often measured in the tens of millions. This makes it extremely challenging to efficiently collect, store, and analyze large numbers of gradients in these models. In this work, we develop MetaStore to fill this gap. MetaStore leverages our observation that storing certain compact intermediate results produced in the back propagation process, namely, the prefix and suffix gradients, is sufficient for the exact restoration of the original gradient. These prefix and suffix gradients are much more compact than the original gradients, thus allowing us to address the gradient collection and storage challenges. Furthermore, MetaStore features a rich set of analytics operators that allow the users to analyze the gradients for data debugging or model interpretation. Rather than first having to restore the original gradients and then run analytics on top of this decompressed view, MetaStore directly executes these operators on the compact prefix and suffix structures, making gradient-based analytics efficient and scalable. Our experiments on popular deep learning models such as VGG, BERT, and ResNet and benchmark image and text datasets demonstrate that MetaStore outperforms strong baseline methods from 4 to 678x in storage costs and from 2 to 1000x in running time. 

A new friction device using band brake technology, termed the Banded Rotary Friction Damper (BRFD), has been fabricated at the NHERI Lehigh Experimental Facility. The damping mechanism is based on band brake technology and leverages a self-energizing mechanism to produce large damping forces with low input energy. The device is a second-generation BRFD, where the friction mechanism is achieved using two electric actuators. The BRFD generates a damping force as a function of the input force provided by the electric actuators, where the ratio of BRFD force output-to-electric actuator force input is equal to about 112. The paper presents the results of a study using real-time hybrid simulations (RTHS) to investigate the performance of the BRFD’s in mitigating seismic hazards of a two-story reinforced concrete building. The building has two and three special moment resisting frames (SMRFs) in the east-west and north-south directions, respectively. In order to perform the RTHS, the north south SMRF is considered and the BRFD along with a parallel elastic member is used as a base isolation system to mitigate the effects of earthquake hazards by reducing story drift and floor accelerations of the structure. For the RTHS the building and the elastic component of the isolator are part of the analytical substructure while the experimental substructure is comprised of the BRFD. The response of the structure is investigated involving six Maximum Considered Earthquake (MCE) hazard level events that includes three near-field and three far-field ground motions. The explicit, unconditionally stable dissipative Modified KR-α integration algorithm is used to accurately integrate the equations of motion during the RTHS. The model for the reinforced concrete building is created using explicit non-linear force-based fiber elements to discretely model each member of the structure. First, the details of the prototype of the BRFD are presented. Second, the details of the isolator system consisting of a linear spring element and the BRFD are discussed. Finally, the details of the RTHS study and the results are presented. The building’s inter-story peak and residual story drift from base-isolated and fixed-based conditions are compared. Results show that the proposed isolator system produces a significant reduction in both maximum inter-story drift and residual drift, and reduces the damage developed in the structure during the MCE. 

In this paper, we present convergence guarantees for a modified trust-region method designed for minimizing objective functions whose value and gradient and Hessian estimates are computed with noise. These estimates are produced by generic stochastic oracles, which are not assumed to be unbiased or consistent. We introduce these oracles and show that they are more general and have more relaxed assumptions than the stochastic oracles used in prior literature on stochastic trust-region methods. Our method utilizes a relaxed step acceptance criterion and a cautious trust-region radius updating strategy which allows us to derive exponentially decaying tail bounds on the iteration complexity for convergence to points that satisfy approximate first- and second-order optimality conditions. Finally, we present two sets of numerical results. We first explore the tightness of our theoretical results on an example with adversarial zeroth- and first-order oracles. We then investigate the performance of the modified trust-region algorithm on standard noisy derivative-free optimization problems.