20005 research outputs found
Sort by
Deep eutectic solvent pre-treatment of residual biomass streams – Effects on anaerobic degradability
No full textThe suitability of residual biomass from biorefineries as biogas substrate, including the effect of DES pre-treatment on biogas potential and biomass composition was evaluated. Three different biomasses, i.e., cork dust, olive tree pruning, and common reed, were treated with a deep eutectic solvent (DES; i.e., choline chloride and formic acid, 1:2 M ratio, 1:10 (g mL−1) solid to solvent ratio) under various conditions. While being ineffective under less severe conditions, significantly increased biogas potentials were observed for all biomasses at 130 °C and 40 and/or 60 min of pre-treatment. Cork dust had the highest relative increase in biogas potential of 125 % (130 °C, 60 min). With an increase of 90 mLN gVS−1 (+21 %), olive tree pruning showed the highest absolute increase. Common reed demonstrated a notable increase of 12 %. Concurrently, kinetic modelling revealed shifts in degradation kinetics, such as altered lag phases and higher maximum biogas rates, hereby the Cone Model provided the best fit for the data. The change could be attributed to delignification and retention of the fermentable glucan fraction in the solid biomass. Despite considerable differences among the biomasses, the results underscore the efficacy of DES pre-treatment in enhancing the anaerobic degradability of residual biomass
Enhancing Practicality of Memory Compression for GPUs with High-Throughput Simplifications
No full textMemory-bound Graphics Processing Unit (GPU) applications are limited by memory bandwidth, as the rapid growth in computational power has outpaced the slower increase in memory bandwidth. Consequently, approaches such as memory compression, are gaining prominence to synthetically enhance memory bandwidth and accelerate bandwidth-limited applications. Traditionally, compression techniques have been tailored towards achieving high compression ratios without considering the high bandwidth of modern GPU memory systems which makes hardware integration costly and impractical. We analyze several state-of-the-art memory compression techniques and finds that the throughput of Bit-Plane Compression (BPC) and Frequent Pattern Compression (FPC) is limited by Zero Run-Length Encoding (ZRLE), which efficiently compresses zero blocks, however, GPUs often do not benefit as even heavily compressed blocks require the transfer of a full Memory-Access Granularity (MAG). We propose simplifying the BPC and FPC techniques by removing ZRLE and introducing a fixed-size tag section. Together with higher word-level parallelism, our simplifications increase compressor throughput by 14.0× and decompressor throughput by 13.5x without any loss in the effective compression ratio. Additionally, the area required for hardware integration is significantly reduced; for instance, the area cost of BPC is decreased by 3.6x, and power consumption by 1.8x, making hardware integration of memory compression more practical and cost-effective
Machine learning pipeline for structure–property modeling in Mg-alloys using microstructure and texture descriptors
No full textIdentifying the relationships between material structure and mechanical properties has been crucial for accelerating the exploration of the material design space for advanced alloys. However, traditional approaches for magnesium (Mg) alloys often fall short in providing quantitative and broadly applicable structure–property linkages. To address this challenge, a comprehensive machine learning pipeline is presented for structure–property modeling in extruded Mg-alloys, leveraging both microstructure and texture descriptors derived from experimental data. The pipeline encompasses a robust workflow for data extraction from optical microscopy and X-ray diffraction, advanced image processing and deep learning techniques for microstructure binarization and grain statistics, and the computation of statistical descriptors including n-point spatial correlations, gram matrices for microstructure, and generalized spherical harmonics (GSH) for texture. Dimensionality reduction techniques such as principal component analysis (PCA), isomap, and autoencoders are employed to manage the high-dimensionality of the descriptor space. Subsequently, non-linear regression models—Gaussian Process, XGBoost, and Multi-Layer Perceptron regressors—are evaluated to predict mechanical properties, specifically strain hardening exponent (n) and yield stress (σy). Our results demonstrate that XGBoost consistently outperforms other regressors, achieving a notably low mean absolute percentage error (MAPE) of 6.67% for strain hardening exponent and 7.01% for yield stress, using a combination of PCA-reduced 3-point spatial correlations and isomap-reduced gram matrices as microstructure descriptors, and isomap-reduced GSH coefficients as texture descriptors at a 150μm length scale. Shapley Additive exPlanations (SHAP) analysis further reveals that texture descriptors and aspect ratio distribution are the most influential features in predicting mechanical properties. This established ML framework for structure–property modeling in Mg-alloys, surpasses state-of-the-art benchmarks and provides a valuable template for materials design and discovery
Universitätsbibliothek TU Hamburg - Jahresbericht 2024
No full textBericht über Services und Entwicklungen und die Kennzahlen der Universitätsbibliothek an der Technischen Universität Hamburg für 2024.Report on services and developments and the key figures of the University Library at Hamburg University of Technology for 2024
BIM-based human-robot collaboration for building inspections using mixed reality
No full textHuman-robot collaboration (HRC) may be leveraged to improve building inspections by allowing robots and humans to perform inspections jointly. However, HRC methods lack intuitive interfaces for seamless collaboration and contextual building information exchange, limiting collaboration and inspection efficiency. This paper proposes a BIM-based HRC framework using mixed reality (MR), enabling operators to visualize inspection data, control robots, and to monitor inspection processes in real time. The BIM-based HRC framework is implemented using MR headsets and validated through collaborative inspections of an indoor environment, showcasing the potential of the framework to advance collaborative building inspections and improve inspection accuracy and efficiency
#WeAreNotWaiting: Relational agency in human-algorithm interaction in an artificial Pancreas system
No full textA burgeoning literature scrutinizes the interactions between humans and algorithms and how agency is (re)distributed between the two parties in the interaction. However, much of this work theorizes two-sided interactions, overlooking multi-agent settings where numerous humans and multiple algorithms may be involved. To understand the redistribution of agency in human-algorithm interaction in multi-agent settings, we conducted a qualitative case study of an open-source artificial pancreas system, in which individuals with Type 1 diabetes interact with other human and nonhuman actors to manage their medical condition. Using a relational agency lens, we unearth the detailed practices in multi-agent human-algorithm interaction to create an initial typology of practices that depict the complexity of collective agency dynamics in multi-agent human-algorithm interaction
Complexity reduction for TSO-DSO coordination: flexibility aggregation vs. distributed optimization
No full textThe increasing number of flexible devices and distributed energy resources in power grids renders the coordination of transmission and distribution systems increasingly complex. In this paper, we discuss and compare two different approaches to optimization-based complexity reduction: Flexibility aggregation via Approximate Dynamic Programming (ADP) and distributed optimization via the Alternating Direction Method of Multipliers (ADMM). Flexibility aggregation achieves near-optimal solutions with minimal communication. However, its performance depends on the quality of the approximation used. In contrast, ADMM attains results closer to the centralized solution but requires significantly more communication steps. We draw upon a case study combining different matpower benchmarks to compare both methods
Large problems are not necessarily hard: a case study on distributed NMPC paying off
No full textA key motivation in the development of Distributed Model Predictive Control (DMPC) is to accelerate centralized Model Predictive Control (MPC) for large-scale systems. DMPC has the prospect of scaling well by parallelizing computations among subsystems. However, communication delays may deteriorate the performance of decentralized optimization, if excessively many iterations are required per control step. Moreover, centralized solvers often exhibit faster asymptotic convergence rates and, by parallelizing costly linear algebra operations, they can also benefit from modern multicore computing architectures. On this canvas, we study the computational performance of cooperative DMPC for linear and nonlinear systems. To this end, we apply a tailored decentralized real-time iteration scheme to frequency control for power systems. DMPC scales well for the considered linear and nonlinear benchmarks, as the iteration number does not depend on the number of subsystems. Comparisons with multi-threaded
centralized solvers demonstrate competitive performance of the proposed decentralized optimization algorithms
Leveraging generative AI tools for design method support: insights, challenges, and best practices
No full textPublicly available generative AI tools, such as ChatGPT, Midjourney, and DALL-E 3, have the potential to transform product development by accelerating tasks and improving design ideation. Through case studies of scenario management and persona storyboarding, this research explores the strengths and limitations of generative AI (GenAI) tools. The results highlight GenAI's ability to accelerate routine tasks, improve ideation, and support iterative design, but also reveal limitations in contextual understanding and output quality. Key findings show that effective GenAI integration depends on precise prompt design, iterative interaction and critical validation. Despite their potential, GenAI tools cannot replace human expertise for nuanced design tasks. The study provides actionable insights and best practices for leveraging GenAI tools, paving the way for enhanced human-AI collaboration
Data-driven and learning-based control – perspectives and prospects
No full textThe use of data to build models for feedback design is a classic topic of systems and control. Indeed, the term system identification can be traced back to the early 1956 work of Lotfi Zadeh [1]. Step and impulse responses remain pivotal concepts in undergraduate control education. Yet, recently, there is a rapidly growing trend towards data-driven and learning-based control in which classic results are revisited and expanded. As a prime example one may consider the 2005 note by Jan C. Willems and co-authors on persistency of excitation [2]. As illustrated in Figure 1, for the first 15 years after its publication this paper did not spark major interest beyond the community working on subspace identification of dynamic systems. Beginning with the understanding that its main result – which is also coined Willems’ fundamental lemma – enables data-driven control of LTI systems, its citation impact bifurcated to exponential growth after 2018. On this canvas, this special issue presents a snapshot of research on data-driven and learning-based methods in the German control community. Inter alia this special issues demonstrates that the ever-growing impact of data-driven and learning-based methods on systems and control goes far beyond Willems’ lemma. This includes papers working also with Koopman operators, neural networks and Gaussian processes. A major commonality of all contributed papers is the consideration of optimization-based approaches. Following the established structure of the journal the articles of this special issue are clustered into two categories – methods and applications