Sharif Digital Repository

Transformer-based language models have shown an excellent ability to effectively capture and utilize contextual information. Although various analysis techniques have been used to quantify and trace the contribution of single contextual cues to a target task such as subject-verb agreement or coreference resolution, scenarios in which multiple relevant cues are available in the context remain underexplored. In this paper, we investigate how language models handle gender agreement when multiple gender cue words are present, each capable of independently disambiguating a target gender pronoun. We analyze two widely used Transformer-based models: BERT, an encoder-based, and GPT-2, a... 

In this paper, the effect of various arrangements of displacement, velocity and acceleration related weighting matrices on the performance of active control systems on nonlinear frames has been studied. Different arrangements of weighting matrices and feedback combinations of the response have been considered to design the active controllers using a single actuator for reducing the response of an eight-storey bilinear hysteretic frame under white noise excitations. The nonlinear Newmark-based instantaneous optimal control algorithm has been used, where the distributed genetic algorithm (DGA) is employed to determine the proper set of weighting matrices. For each set of feedback and weighting... 

In this paper, the problem of coarse timing synchronization in orthogonal frequency-division multiplexing (OFDM) systems is investigated, and two new timing metrics with better performance are presented. The new metrics take advantage of two novel differential normalization functions that are based on the fourth-order statistics and are designed depending on the value of carrier frequency offset (CFO). The proposed timing metrics are theoretically evaluated using two different class separability criteria. It is shown that the new schemes considerably increase the difference between the correct and wrong timing points in comparison with previous methods. The computational complexity of the... 

Due to role of polymer in increasing sweep efficiency during oil recovery, much attention has been paid to the using polymer solutions in enhanced oil recovery methods. In spite of the existence of the great researches in this area, the role of nanoparticles in modification of the polymer performance in the presence of salts has not been examined before. Furthermore, there is no information about how the dispersed silica nanoparticles affect the heavy oil recovery during the polymer flooding in the presence of divalent cations. In this study, a series of polymer flooding experiments are performed in a quarter five-spot glass micromodel saturated with heavy oil. Solutions of polyacrylamide... 

The novel N2O2 unsymmetrical tetradentate Schiff bases complexes are used as catalysts for the selective aerobic oxidation of cyclohexene. In general, conversion percentages decrease with increasing ΔEp and decreasing Eredox 0′. Catalytic activity increases with decrease of the number of electron-donating groups and the catalytic selectivity varies according to the types of substituents in the ligands. The catalytic system described here is an inexpensive method for selective oxidation of olefins, with advantages of high activity, selectivity, re-usability and short reaction times. Analysing GC and redox potential results shows a moderate correlation between selectivity and activity. © 2004... 

Commercially pure aluminum sheets, AA 1050, are processed by accumulative roll bonding (ARB) up to eight cycles to achieve ultrafine-grained (UFG) aluminum as primary material for mechanical testing. Optical microscopy and electron backscattering diffraction analysis are used for microstructural analysis of the processed sheets. Strain rate sensitivity (m-value) of the specimens is measured over a wide range of strain rates by stress relaxation test under plane strain compression. It is shown that the flow stress activation volume is reduced by decrease of the grain size. This reduction which follows a linear relation for UFG specimens, is thought to enhance the required effective (or... 

In this paper, the capabilities of tuned mass dampers (TMDs) for the mitigation of response of nonlinear frame structures subjected to earthquakes have been studied. To determine the optimal parameters of a TMD, including its mass, stiffness and damping, we developed an optimization algorithm based on the minimization of a performance index, defined as a function of the response of the nonlinear structure to be controlled. Distributed genetic algorithm has been used to solve the optimization problem. For illustration, the method has been applied to the design of a linear TMD for an eight-story nonlinear shear building with bilinear hysteretic material behavior subjected to earthquake. The... 

In this paper, optimal design and assessment of the capability of Tuned Mass Dampers (TMDs) in improving the seismic behavior of Confined masonry walls, as the main element of historical buildings, has been studied. For this purpose, the design parameters of TMDs have been determined through minimization of wall response using Genetic Algorithms (GAs). To simulate the behavior of Confined masonry wall under earthquake, the triple linear shear beam model has been used. For illustration, the method has been tested on Confined masonry walls equipped with linear TMDs. To study the effects of frequency content and Peak Ground Acceleration (PGA) of earthquake records on the performance of TMDs,... 

Microbial enhanced oil recovery (MEOR) is receiving renewed interest worldwide in recent years as a viable method while not damaging the reservoir is proven to be remarkably effective, however to some extent costly. This method is based on microorganisms' activities to reduce residual oil of reservoirs, which is dependent on behavior of inherent microorganisms or injection of bioproduct of external microorganisms. In this work, five bacterial species were taken from MIS crude oil that is one of the aging Persian fractured reservoirs. These microorganisms are substantially strong in increasing oil recovery especially by reducing IFT and other MEOR mechanisms such as change of wettability of... 

Microbial EOR (MEOR) is receiving renewed interest worldwide. The method is based on microorganisms' activities to reduce residual oil of reservoirs, which is dependent on behavior of inherent microorganisms or injection of bioproduct of external microorganisms. Five bacterial species were taken from MIS crude oil that is one of the aging Persian fractured reservoirs. Visualization experiments were carried out to examine the behavior of MEOR in micromodels designed to resemble the fractured system: static and dynamic. A new surface sample of the MIS oil was taken at the well head and physical specifications were measured at the lab. This was fairly light oil with API gravity of 42.5. The... 

A new model is proposed to predict the Forming Limit Diagrams (FLDs) of Tailor Welded Blanks (TWBs). This model is derived by modification of the Marciniak and Kuczynski (MK) analysis and imposing the weld constraints to deformation state of each blank at each strain increment. Based on this model, it is possible to predict the strains of both weaker and stronger blanks once necking occurs. By comparison to the experimental results of two laser welded TWBs, it is demonstrated that this model provides a close prediction of FLD of TWBs, when necking occurs far from the weld line at a region out of HAZ (Heat Affected Zone) and close to it so that it is affected by the weld line constraints  

The aim of this work is to produce two layered thin-walled Cu/Al composite tube by the spin-bonding process. The process is utilized to bond the aluminum tube into the copper one at thickness reductions of 20-60% and process temperatures of 25°C, 130°C, and 230°C. The bond strength is measured by T-peeling test, and the bond interfaces are examined by metallography and scanning electron microscopy (SEM). The results show that after a threshold thickness reduction of about 30%, the bond strength increased with the amount of deformation. SEM fractography of the peel surfaces confirms that the copper oxide film is broken in a shear manner during deformation. Severe shear strains applied during... 

A novel SPD process for manufacturing of high strength tubes and cylinders by accumulative spin-bonding (ASB) is proposed. It is demonstrated that due to incremental deformation in this process, high strain rate without considerable temperature rise is achieved. This is accompanied with a high value of Zener-Hollomon parameter as a characteristic of this SPD process. ASB was applied to a commercially pure aluminum up to four cycles and its effects on the microstructure and mechanical properties were examined by optical microscopy, TEM, EBSD, microhardness and tension tests. The results show that ultra-fine grains are developed during the process by formation of subgrains at early stages... 

Utilizing the so-called tube spinning, a cold-bonding process entitled as "spin-bonding" is developed to produce seamless thin-walled clad tubes and cylinders. By this method, two layers of AA 1050 tubes were successfully bonded together to form a clad tube at room temperature. Based on mechanical aspects of the tube spinning process, the mechanism of spin-bonding is explained in two stages: surface preparation before occurrence of a stable bond and bond strengthening thereafter. The effects of process temperature, thickness reduction, feed rate and roller attack angle as the parameters of tube spinning on the bond strength are studied. It is shown that the bond strength increases by... 

Pain level estimation from videos of face has many benefits for clinical applications. Most of the previous works focused only on pain detection task. However, pain level estimation of video sequences has been discussed fewer. In this work, we have proposed a new regression-based approach to estimate the pain level of video sequences. As the first step, facial expression-related features were extracted from each frame, this task was done by reducing identity-related features using the robust principal component analysis decomposition. Then, we used the minimum, maximum, and mean of the features of frames in a sequence to represent that sequence by a fixed-length feature vector. After this,... 

In this study, ultrafine grained aluminum sheets AA1050 and AA5052 are incrementally cryo-rolled after precooling in liquid nitrogen. Cryogenic deformation is considered as an effective approach to postpone saturation of the grain refinement by conventional severe plastic deformation. However, for a quantitative understanding of the resulted microstructure and mechanical properties, accurate values of the deformation parameters (strain, strain rate and temperature) are needed. The aim of this study is, therefore, to find values of the applied cryo-rolling parameters. As the experimental measurements of parameters such as temperature are infeasible in the thin specimens, a FEM simulation is... 

Strain rate sensitivity (m-value) of ultrafine grain (UFG) AA 1050 and AA 5052 sheets processed by accumulative roll-bonding is investigated versus strain rate by stress relaxation (SR) test at ambient temperature. The results show a weak viscous nature of deformation for AA 5052 specimens as compared to AA 1050 ones. So that much less stress relaxation and negligible strain rate sensitivity are obtained for this material due to dislocation and grain boundary mobility limitation caused by Mg solute atoms. In order to formulate strain rate sensitivity of UFG aluminum as a function of strain rate, three phenomenological and two empirical models are developed and assessed by the experimental... 

In this paper a new method of modeling shear force-displacement relationship for confined masonry walls by neural networks has been presented. Although the mathematical models have been very useful in the simulations so far, however developing more accurate models is necessary. While developing precise mathematical models for highly hysteretic materials is itself challenging and practically cumbersome, the use of learning algorithms is an attractive alternative. The issue of material modeling by neural networks has been a challenging one itself, noticing available neural networks have some limitations in the learning of non-linearity. In this paper a new type neural network, called Prandtl...