Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
Regional seismic fragility analysis of building clusters considering site–city interactions
No full textWith the continued expansion of urban development, many built-up areas have become increasingly dense. Under earthquake excitations, the seismic structural responses can be mutually affected through interactions between the buildings and the underlying site—commonly referred to as site–city interactions (SCI). However, the limited understanding on SCI effects has hindered their integration into city- or community-scale earthquake risk analyses and disaster mitigation. This study incorporates a numerical coupling approach within a probabilistic framework to evaluate the seismic fragility of more than 100 buildings in the Shanghai CBD while accounting for SCI. A systematic comparison is carried out among three models: (i) no inter-building interactions, (ii) interactions among all buildings, and (iii) interactions among all buildings except the tallest one. The results indicate that, SCI exerts a notable influence on the exceedance probabilities of building damage, whereas the effects on the mean (average) fragility intensity are less pronounced. This suggests that the impact of SCI may be underestimated in scenario-specific assessments. Moreover, the presence of the tallest building increases the damage exceedance probabilities of adjacent structures by over 15 %, highlighting the significant role of extreme-height structures. Overall, these findings underscore the necessity of incorporating site–city interactions into seismic risk assessments to improve urban disaster prevention strategies.</p
A novel design of Transformer Encoder-only network model for fault diagnosis in strongly coupled industrial processes
No full textTo address the impact of strong coupling in chemical processes fault diagnosis, this paper proposes a novel fault diagnosis network called EOI: Encoder for “One-dimensional Image.” By modifying the Transformer Encoder-only architecture to construct the EOI network, we enable the computer to understand this “image” through supervised learning. The architecture proposed in this paper mainly contains three improvements to the Encoder. Firstly, it removes the position encoding to prevent the introduction of noise signals. Secondly, it employs the Squeeze-and-Excitation (SE) attention module from the field of image processing and adapts it to a one-dimensional version. Finally, it improves the traditional Feed-Forward Network (FFN) module based on the idea of Convolutional Neural Network (CNN). The designed EOI network integrates traditional CNN and Long Short-Term Memory (LSTM) networks, significantly enhancing feature extraction efficiency for chemical time-series data and improving fault diagnosis accuracy. Experiments on the Tennessee Eastman (TE) process and industrial coke furnace have demonstrated that the proposed network exhibits high performance.</p
The Impact of Bank Financing on Borrowers’ Voluntary Disclosures and Real Investments
No full textThis paper examines how an increase in bank lending affects firms’ investment disclosures and policies. Exploiting the unconventional liquidity injections by the European Central Bank in 2011–2012, we find that U.S. borrowers of European Union (EU) banks receive greater bank financing and issue more management capital expenditure (capex) forecasts following the liquidity injections. In contrast, we find no evidence of changes in bond financing or issuance of management earnings forecasts. We also find that EU banks’ U.S. borrowers on average receive negative market reactions to their capex forecasts and make no change in investment following the liquidity injection. Further, relative to firms with positive feedback, firms with negative feedback have a lower increase in investment but a higher increase in financial assets. Additionally, we find that the effect of increased bank lending on management capex forecasts is more pronounced among banks with weaker financial conditions and among borrowers with greater financial constraints, higher growth volatility, and more informed trading. These findings highlight the important role of capex forecasts in seeking market feedback following increased bank lending and suggest that market feedback helps explain why the transmission of bank credit to corporate investment did not occur following the liquidity injections.</p
A Rolling Recruitment Process Under Applicant Stochastic Departures
No full textProblem definition: We study a rolling recruitment process in which applicants leave the system stochastically. Applicants arrive randomly over time, and each applicant is available for a random amount of time after they arrive. In each period, the recruiter must decide whether to stop or to wait. If the recruiter stops, he or she needs to determine how many offers to make and whom to make offers to, and the applicants who do not receive an offer will leave the system. If the recruiter waits, more applicants will be available in the next period, whereas some who arrived earlier will leave. Methodology/results: We model the process as a large-scale optimal stopping problem and show how the applicant qualifications, measured by scores, affect the recruiter’s optimal policy. Managerial implications: We find that the optimal stopping rule for each applicant’s score is a two-threshold policy. If the score exceeds the higher threshold, then the recruiter stops and makes an offer to the applicant and possibly to others. If the score falls below the lower threshold, then the recruiter also stops but makes no offer to the applicant. If the score is in between the two thresholds, the recruiter waits. We further explore the impact on an applicant’s likelihood of receiving an offer if his or her competitors become more qualified. When the score of another applicant increases, the recruiter may change from making an offer to an applicant to waiting or to instead making an offer to the other applicant whose score has increased. In other words, an applicant may be disadvantaged if he or she faces stronger competitors, which is expected. However, an applicant may also benefit from having stronger competitors. When the score of another applicant increases, the recruiter may change from not making an offer to an applicant to making an offer to both applicants. Overall, we provide valuable insights into the role of applicant qualifications in stopping decisions, propose methods for computing the optimal policy, and quantify the benefits of endogenously determining the stopping rule.</p
A Primal-Dual Approach to Constrained Markov Decision Processes with Applications to Queue Scheduling and Inventory Management
No full textIn many operations management problems, we need to make decisions sequentially to minimize the cost, satisfying certain constraints. One modeling approach to such problems is the constrained Markov decision process (CMDP). In this work, we develop a data-driven primal-dual algorithm to solve CMDPs. Our approach alternatively applies regularized policy iteration to improve the policy and subgradient ascent to maintain the constraints. Under mild regularity conditions, we show that the algorithm converges at rate (Formula Presented), where T is the number of iterations, for both the discounted and long-run average cost formulations. Our algorithm can be easily combined with advanced deep learning techniques to deal with complex large-scale problems with the additional benefit of straightforward convergence analysis. When the CMDP has a weakly coupled structure, our approach can further reduce the computational complexity through an embedded decomposition. We apply the algorithm to two operations management problems: multiclass queue scheduling and multiproduct inventory management. Numerical experiments demonstrate that our algorithm, when combined with appropriate value function approximations, generates policies that achieve superior performance compared with state-of-the-art heuristics.</p
Valley-polarized Josephson junctions as gate-tunable 0-π qubit platforms
No full textRecently, gate-defined Josephson junctions based on magic-angle twisted bilayer graphene (MATBG) have been fabricated. In such a junction, local electrostatic gating can create two superconducting regions connected by an interaction-driven valley-polarized state as the weak link. Due to the spontaneous time-reversal and inversion symmetry breaking, novel phenomena such as the Josephson diode effect have been observed with zero external magnetic fields. Importantly, when the so-called nonreciprocity efficiency (which measures the sign and strength of the Josephson diode effect) changes sign, the energy-phase relation of the junction is approximately F(ϕ)∝cos(2ϕ) where F is the free energy and ϕ is the phase difference of the two superconductors. In this work, we show that such a MATBG-based Josephson junction, when shunted by a capacitor, can be used to realize the long-sought-after 0-π qubits which are protected from local perturbation-induced decoherence. Interestingly, by changing the junction parameters to the regime where a large nonreciprocity efficiency is obtainable, transmon-like qubits with large anharmonicity can also be realized. The gate-defined Josephson junctions can be employed as platforms for realizing qubits that are protected from local perturbations.</p
Biological Effects of Toluene Exposure on A549 Cells under Air–Liquid Interface Conditions
No full text
