48 research outputs found
Reliability-based co-design and its applications to wind energy and mobile energy storage systems
Get PDFAutonomous systems, such as autonomous driving vehicles, unmanned aerial vehicles (UAVs), and field robots, received much attentions recently. The performance of autonomous systems relies on both its physical design and the appropriate control strategies, which often takes place at an early stage of design. The plant design and the control design are strongly coupled. Neglecting this coupling effect may cause an imbalance in the feasible design spaces of plant design and control design, such as over-constrained operation conditions, over design, or requirement of skilled operators, which hinders the development of autonomous systems. On the other hand, the products are manufactured goods and usually operate in environments with uncertainty. Reliable operation of such systems ask for balanced physical design and feasible control decisions to address the parametric uncertainty and stochastic environmental disturbances.
While integrated physical and control system co-design has been demonstrated successfully on several engineering system design applications, it has been primarily applied in a deterministic manner without considering uncertainties. An opportunity exists to study non-deterministic co-design strategies, taking into account various uncertainties in an integrated co-design framework. While significant advancements have been made in co-design and RBDO separately, little is known about methods where reliability-based dynamic system design and control design optimization are considered jointly. In this research, we investigate optimal design and control of dynamical systems with model parametric uncertainties, which presumably operate in uncertain environments. Techniques in control co-design (CCD) and reliability-based design optimization (RBDO) are adapted and integrated to solve the proposed problem. Since the proposed method adopts the idea of multi-disciplinary design optimization, it can improve the performance of autonomous systems without leveraging the difficulty in design and control for systems with uncertainties.
First, the problem formulation and strategies to solve the reliability-based control co-design problem is presented. A comparison of accuracy and efficiency is made using numerical and simple engineering case studies. The method is then applied to a horizontal axis wind turbine. The uncertain wind load and model parameters of a wind turbine are compensated through active control or endured by a reliable design regarding its aerodynamics and structural dynamics. Different strategies of reliability assessment are also compared, which provides insights on their advantages and limits under different cases.
In the second application, reliability-based control co-design is applied to Lithium-ion battery. The electrode and charging current are optimized to minimize its charging time while regulating its aging effect for reasonable cycle life. The multi-scale nature of the problem requires first principle model to preserve the coupling effect between electrode design at the micro scale and the charging control at the macro scale. However, it is not feasible to use the first principle model for control optimization. A hybrid physics and machine learning strategy is proposed in this work, which extends the applicability of reliability-based control co-design to multi-scale problems.Submission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2023-08-01The student, Tonghui Cui, accepted the attached license on 2021-07-14 at 12:23.The student, Tonghui Cui, submitted this Dissertation for approval on 2021-07-14 at 12:46.This Dissertation was approved for publication on 2021-07-16 at 14:17.DSpace SAF Submission Ingestion Package generated from Vireo submission #16820 on 2022-01-12 at 13:04:15Made available in DSpace on 2022-01-12T22:55:01Z (GMT). No. of bitstreams: 2
CUI-DISSERTATION-2021.pdf: 6407126 bytes, checksum: 20e098cc18731c293e76aeab7fd810b9 (MD5)
LICENSE.txt: 4208 bytes, checksum: 158d4051d87a7f57d8452147e3172ad5 (MD5)
Previous issue date: 2021-07-16Embargo set by: Seth Robbins for item 121217
Lift date: 2024-01-12T22:55:09Z
Reason: Author requested closed access (OA after 2yrs) in Vireo ETD systemEmbargo set by: Seth Robbins for item 121217
Lift date: 2024-01-12T22:56:20Z
Reason: Author requested closed access (OA after 2yrs) in Vireo ETD systemAuthor requested closed access (OA after 2yrs) in Vireo ETD systemLimite
Multi-channel quasi-adaptive processing based on low and slow targets in complex environment
No full textLate Ordovician to early Devonian adakites and Nb-enriched basalts in the Liuyuan area, Beishan, NW China: Implications for early Paleozoic slab-melting and crustal growth in the southern Altaids
No full textWe report newly-defined Nb-enriched basalts, adakites and dacites from the Beishan, NW China of the southern Altaids based on field, geochemical, isotopic and geochronology studies. Two phases of adakites (adakite-I and adakite-II) have been defined, which are calc-alkaline, and characterized by high Na 2O/K 2O ratios (1.49-1.71 and 2.32-3.64) and Sr contents (494-1213 ppm and 325-494 ppm), negligible to positive Eu anomalies, strong depletion of HREE (e.g., Yb = 0.48-0.93 ppm and 0.50-0.99 ppm) and Y (6.87-9.80 ppm and 6.02-10.30 ppm), and enriched in Rb, Sr, Ba, K and depleted Nb and Ti. They are characterized by relatively low ε Nd(t) values (- 0.8 to - 0.9 and + 0.6 to + 3.8) and relatively constant high ( 87Sr/ 86Sr) i ratios (0.70635-0.70636 and 0.70583-0.70651). The zircons of adakite-I have relatively low ε Hf(t)(- 0.8 to + 2.7). The Nb-enriched basalts are sodium-rich (N 2O/K 2O = 1.31-4.44), with higher TiO 2, P 2O 5, Zr and Nb contents and (Nb/Th) PM, (Nb/La) PM and Nb/U ratios than typical arc basalts. They are relatively enriched in Rb, Ba, U, Pb and K, depleted in Nb, and minor negative to positive Ba, Zr, Sr and Ti. They have low positive ε Nd(t) (+ 0.9 to + 2.3) and relatively high ( 87Sr/ 86Sr) i (0.70556-0.70691) ratios. The dacites are typical arc magmas, with moderately enriched LILE, distinctly negative Eu, Nb, Sr and Ti anomalies. They have positive ε Nd(t) (+ 2.2) and relatively high ( 87Sr/ 86Sr) i (0.70786). We argue that the Liuyuan adakites were most probably related to the melting of young/hot subducted crust of the Paleo-Asian Ocean, which included tectonically-subducted radiogenic crustal material and/or inheritance from highly radiogenic oceanic crust (e.g. OIB). The Nb-enriched basalts likely resulted from mantle wedge peridotites metasomatized by adakites and/or further changed by components other than adakites (e.g., minor slab-derived fluids). Based on own zircon SIMS U-Pb dating of these key rock types, we further propose that from the late Ordovician to early Devonian, large volumes of magma consisting of late Ordovician Nb-enriched basalts (451 Ma) and dacites (442 Ma), late Silurian adakite-I (424 Ma), early Devonian adakite-II (374 Ma) and I-S-A-type granites (436 Ma-380 Ma), developed in the southern Altaids. Together with other geochronological data from the literature, we conclude that subducted oceanic slab-melting was frequent from 470 Ma to 370 Ma. Our results suggest that frequent hot (and/or young) oceanic crustal subduction and slab-melting were important mechanisms in the accretionary growth of the Southern Altaids. © 2011 International Association for Gondwana Research.link_to_subscribed_fulltex
Significant role of ultramicropores on capacitive properties of polypyrrole-based carbons
No full text
From Complex Word Identification to Substitution:Instruction-Tuned Language Models for Lexical Simplification
Get PDFLexical-level sentence simplification is essential for improving text accessibility, yet traditional methods often struggle to dynamically identify complex terms and generate contextually appropriate substitutions, resulting in limited generalization. While prompt-based approaches with large language models (LLMs) have shown strong performance and adaptability, they often lack interpretability and are prone to hallucinating. This study proposes a fine-tuning approach for mid-sized LLMs to emulate the lexical simplification pipeline. We transform complex word identification datasets into an instruction–response format to support instruction tuning. Experimental results show that our method substantially enhances complex word identification accuracy with reduced hallucinations while achieving competitive performance on lexical simplification benchmarks. Furthermore, we find that integrating fine-tuning with prompt engineering reduces dependency on manual prompt optimization, leading to a more efficient simplification framework
Using Feature Filtering Metrics as Meta-dimensions in Constructing Distributional Representations
No full textAnalysis of community‐level factors contributing to cholera infection and water testing access in the Northern Corridor of Haiti
No full textCan Major Public Health Emergencies Affect Changes in International Oil Prices?
No full textIn order to deepen the understanding of the impact of major public health emergencies on the oil market and to enhance the risk response capability, this study analyzed the logical relationship between major public health emergencies and international oil price changes, identified the change points, and calculated the probability of abrupt changes to international oil prices. Based on monthly data during six major public health emergencies from 2009 to 2020, this study built a product partition model. The results show that only the influenza A (H1N1) and COVID-19 pandemics were significant reasons for abrupt changes in international oil prices. Furthermore, the wild poliovirus epidemic, the Ebola epidemic, the Zika epidemic, and the Ebola epidemic in the Democratic Republic of the Congo had limited effects. Overall, the outbreak of a Public Health Emergency of International Concern (PHEIC) in major global economies has a more pronounced impact on international oil prices
Semantic enrichment of neural word embeddings: Leveraging taxonomic similarity for enhanced distributional semantics
Get PDFData-driven neural word embeddings (NWEs), grounded in distributional semantics, can capture various ranges of linguistic regularities, which can be further enriched by incorporating structured knowledge resources. This work proposes a novel post-processing approach for injecting semantic relationships into the vector space of both static and contextualized NWEs. Current solutions to retrofitting (RF) word embeddings often oversimplify the integration of semantic knowledge, neglecting the nuanced differences between relationships, which may result in suboptimal performance. Instead of applying multi-thresholding to distance boundaries in metric learning, we compute taxonomic similarity to dynamically adjust these boundaries during the semantic specialization of word embeddings. Benchmark evaluations on both static and contextualized word embeddings demonstrate that our dynamic-fitting (DF) approach produces SOTA correlation results of 0.78 and 0.76 on SimLex-999 and SimVerb-3500, respectively, highlighting the effectiveness of incorporating multiple semantic relationships in refining vector semantics. Our approach also outperforms existing RF methods in both supervised and unsupervised semantic relationships recognition tasks. It achieves top accuracy scores for hypernymy detection on the BLESS, WBLESS, and BIBLESS datasets (0.97, 0.89, and 0.83, respectively) and an F1 score of over 0.60 on four types of semantic relationship classification in the shared Subtask-2 of CogALex-V, surpassing all participant systems. In the analogy reasoning task of the Bigger Analogy Test Set, our approach outperforms existing RF methods on inferring relational similarity. These consistent improvements across various lexical semantics tasks suggest that our DF approach can effectively integrate distributional semantics with symbolic knowledge resources, thereby enhancing the representation capacity of word embeddings in downstream applications
