2024 - Repozytorium publikacji - Politechnika Gdańska

Pokaż publikacje z roku

Publikacje z roku 2024

Pokaż wszystkie

An Analysis of the Performance of Lightweight CNNs in the Context of Object Detection on Mobile Phones
- Jakub Łęcki
- Marek Hering
- Maciej Jabłoński
- Aleksandra Karpus
2024
Convolutional Neural Networks (CNNs) are widely used in computer vision, which is now increasingly used in mobile phones. The problem is that smartphones do not have much processing power. Initially, CNNs focused solely on increasing accuracy. High-end computing devices are most often used in this type of research. The most popular application of lightweight CNN object detection is real-time image processing, which can be found in devices such as cameras and autonomous vehicles. Therefore, there is a need to optimize CNNs for use on mobile devices. This paper presents the comparision of latency and mAP of 22 lightweight CNN models from the MobileNet and EfficientDet families measured on 7 mobile phones.
An analytical approach to determine the health benefits and health risks of consuming berry juices
- Magdalena Fabjanowicz
- Anna Rożańska
- Nada S. Abdelwahab
- Marina Pereira-Coelho
- Isabel Cristina da Silva Haas
- Luiz Augusto dos Santos Madureira
- Justyna Płotka-Wasylka
2024 Pełny tekst FOOD CHEMISTRY
Food products composition analysis is a prerequisite for verification of product quality, fulfillment of regulatory enforcements, checking compliance with national and international food standards, contracting specifications, and nutrient labeling requirements and providing quality assurance for use of the product for the supplemen- tation of other foods. These aspects also apply to the berry fruit and berry juice. It also must be noted that even though fruit juices are generally considered healthy, there are many risks associated with mishandling both fruits and juices themselves. The review gathers information related with the health benefits and risk associated with the consumption of berry fruit juices. Moreover, the focus was paid to the quality assurance of berry fruit juice. Thus, the analytical methods used for determination of compounds influencing the sensory and nutritional characteristics of fruit juice as well as potential contaminants or adulterations.
An ANN-Based Method for On-Load Tap Changer Control in LV Networks with a Large Share of Photovoltaics—Comparative Analysis
- Klara Janiga
- Piotr Miller
- Robert Małkowski
- Michał Izdebski
2024 ENERGIES
The paper proposes a new local method of controlling the on-load tap changer (OLTC) of a transformer to mitigate negative voltage phenomena in low-voltage (LV) networks with a high penetration of photovoltaic (PV) installations. The essence of the method is the use of the load compensation (LC) function with settings determined via artificial neural network (ANN) algorithms. The proposed method was compared with other selected local methods recommended in European regulations, in particular with those currently required by Polish distribution system operators (DSOs). Comparative studies were performed using the model of the 116-bus IEEE test network, taking into account the unbalance in the network and the voltage variation on the medium voltage (MV) side.
An annotated timeline of sensitivity analysis
- Marta Kuc-Czarnecka
- Stefano Tarantolo
- Federico Ferretti
- Samuele Lo Piano
- Mariia Kozlova
- Alesio Lachi
- Rosana Rosati
- Arnald Puy,
- Pamphile Roy
- Giulia Vannucci
- Andrea Saltelli,
2024 Pełny tekst ENVIRONMENTAL MODELLING & SOFTWARE
The last half a century has seen spectacular progresses in computing and modelling in a variety of fields, applications, and methodologies. Over the same period, a cross-disciplinary field known as sensitivity analysis has been making its first steps, evolving from the design of experiments for laboratory or field studies, also called ‘in-vivo’, to the so-called experiments ‘in-silico’. Some disciplines were quick to realize the importance of sensitivity analysis, whereas others are still lagging behind. Major tensions within the evolution of this discipline arise from the interplay between local vs global perspectives in the analysis as well as the juxtaposition of the mathematical complexification and the desire for practical applicability. In this work, we retrace these main steps with some attention to the methods and through a bibliometric survey to assess the accomplishments of sensitivity analysis and to identify the potential for its future advancement with a focus on relevant disciplines, such as the environmental field.
An automated learning model for twitter sentiment analysis using Ranger AdaBelief optimizer based Bidirectional Long Short Term Memory
- Sasirekha Natarajan
- Smitha Kurian
- Parameshachari Bidare Divakarachari
- Przemysław Falkowski-Gilski
2024 EXPERT SYSTEMS
Sentiment analysis is an automated approach which is utilized in process of analysing textual data to describe public opinion. The sentiment analysis has major role in creating impact in the day-to-day life of individuals. However, a precise interpretation of text still relies as a major concern in classifying sentiment. So, this research introduced Bidirectional Long Short Term Memory with Ranger AdaBelief Optimizer (Bi-LSTM RAO) to classify sentiment of tweets. Initially, data is obtained from Twitter API, Sentiment 140 and Stanford Sentiment Treebank-2 (SST-2). The raw data is pre-processed and it is subjected to feature extraction which is performed using Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF). The feature selection is performed using Gazelle Optimization Algorithm (GOA) which removes the irrelevant or redundant features that maximized model performance and classification is performed using Bi LSTM–RAO. The RAO optimizes the loss function of Bi-LSTM model that maximized accuracy. The classification accuracy of proposed method for Twitter API, Sentiment 140 and SST 2 dataset is obtained as 909.44%, 99.71% and 99.86%, respectively. These obtained results are comparably higher than ensemble framework, Robustly Optimized BERT and Gated Recurrent Unit (RoBERTa-GRU), Logistic Regression-Long Short Term Memory (LR-LSTM), Convolutional Bi-LSTM, Sentiment and Context Aware Attention-based Hybrid Deep Neural Network (SCA-HDNN) and Stochastic Gradient Descent optimization based Stochastic Gate Neural Network (SGD-SGNN).
An Efficient PEEC-Based Method for Full-Wave Analysis of Microstrip Structures
- Jinyan Ma
- Da Li
- Hanzhi Ma
- Ruifeng Li
- Ling Zhang
- Michał Mrozowski
- Erping Li
2024 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY
This article introduces an efficient method for the equivalent circuit characterization and full-wave analysis of microstrip structures, leveraging the full-wave partial element equivalent circuit (PEEC). In particular, the multilayered Green's function is evaluated using the discrete complex-image method (DCIM) and employed to establish the mixed potential integral equations. The proposed strategy considers time delays for the retarded electric and magnetic couplings, offering a new efficient full-wave approach to extract equivalent circuit components, which encapsulate the contributions of the quasi-static, surface-wave, and complex images. It is noted that the proposed full-wave PEEC strategy allows each component contribution derived from DCIM to be efficiently represented as frequency-independent lumped circuit elements and corresponding frequency factors, thereby simplifying the extraction process of the entire frequency-dependent lumped elements in the traditional PEEC method. Moreover, the proposed PEEC model, equipped with full-wave equivalent circuits, offers clear physical insight into electromagnetic behaviors, thereby facilitating design and optimization. Finally, the accuracy and efficiency of the proposed PEEC model are fully demonstrated through various examples and experiments.
An Empirical Study of a Dynamic Stop Loss Strategy with Deep Reinforcement Learning on the NASDAQ Stock Market
- Mateusz Anders
- Jozef Zurada
- Paweł Weichbroth
2024
The objective of this paper is to empirically investigate the efficacy of using Deep Reinforcement Learning (DRL) to maximize investment returns by incorporating expected optimal closing prices of long positions into a daily strategy. This paper extends existing research on the impact of stop-loss orders on investment strategy results and brings contribution of these orders to trading strategies into a completely new perspective. We propose a novel approach using DRL, in contrast to fixed-price stop-loss strategies, trailing-stop strategies, or other machine learning approaches. In the backtesting experiment, daily OHLCV data for stocks from the NASDAQ-100 index (as of May 2024) were used for the period spanning from January 2014 to January 2024. The strategy is compared with buy-and-hold, stop-loss, and trailing stop-loss strategies. Significant effort was made to accurately reflect market conditions in the simulation. We found a positive impact of using DRL compared to other tested strategies when encountering entirely new data, suggesting positive serial market correlations. The results suggest that appropriate closing rules and active management of stop levels can increase investment returns without necessarily reducing portfolio return volatility.
An Example of Using Low-Cost LiDAR Technology for 3D Modeling and Assessment of Degradation of Heritage Structures and Buildings
- Piotr Kędziorski
- Marcin Jagoda
- Paweł Tysiąc
- Jacek Katzer
2024 Materials
This article examines the potential of low-cost LiDAR technology for 3D modeling and assessment of the degradation of historic buildings, using a section of the Koszalin city walls in Poland as a case study. Traditional terrestrial laser scanning (TLS) offers high accuracy but is expensive. The study assessed whether more accessible LiDAR options, such as those integrated with mobile devices such as the Apple iPad Pro, can serve as viable alternatives. This study was conducted in two phases—first assessing measurement accuracy and then assessing degradation detection—using tools such as the FreeScan Combo scanner and the Z+F 5016 IMAGER TLS. The results show that, while low-cost LiDAR is suitable for small-scale documentation, its accuracy decreases for larger, complex structures compared to TLS. Despite these limitations, this study suggests that low-cost LiDAR can reduce costs and improve access to heritage conservation, although further development of mobile applications is recommended.
An Extremely Compact Frequency Reconfigurable Antenna Diplexer Employing Dielectric Liquids
- Rusan Kumar Barik
- Xiaohu Wu
- Xiaoguang Liu
- Sławomir Kozieł
2024 IEEE Antennas and Wireless Propagation Letters
The letter presents an extremely compact frequency reconfigurable antenna diplexer based on fluidic channels for sub6 GHz applications. The proposed antenna diplexer is modelled by employing half-mode (HM) and quarter-mode (QM) substrateintegrated rectangular cavities, two slots, orthogonal feed lines, and fluidic vias. To comprehend the radiation mechanism, the equivalent circuit, electric field distributions, and frequency responses are analyzed. Utilization of HM and QM cavities that are loaded with slots results in an extremely compact antenna diplexer. Three fluidic vias are bored from the bottom plane of each cavity and filled with different dielectric liquids to enable frequency reconfigurability. For validation of the concept, an antenna diplexer is built and demonstrated. The constructed antenna prototype has a small footprint 0.078lg2 with 15% and 16% of reconfigurability in lower and upper frequency bands, respectively. The proposed antenna offers high-isolation exceeding 28 dB, realized gain better than > 3.8 dBi, front-to-back-ratio of > −18 dB, and cross-polarization level of > −18 dB. A good consistency is obtained in between full-wave simulations and measurement.
An image processing approach for fatigue crack identification in cellulose acetate replicas
- Krzysztof Pałczyński
- Jan Seyda
- Dariusz Skibicki
- Łukasz Pejkowski
- Wojciech Macek
2024 ENGINEERING FAILURE ANALYSIS
The cellulose acetate replication technique is an important method for studying material fatigue. However, extracting accurate information from pictures of cellulose replicas poses challenges because of distortions and numerous artifacts. This paper presents an image processing procedure for effective fatigue crack identification in plastic replicas. The approach employs thresholding, adaptive Gaussian thresholding, and Otsu binarization to convert gray-scale images into binary ones, enhancing crack visibility. Morphological operations refine object shapes, and Connected Components Analysis facilitates crack identification. Despite limited data, the handcrafted feature extraction algorithm proves robust, addressing challenges. The algorithm shows efficacy in detecting cracks as small as 30 μm, even in the presence of cellulose replication artifacts. The results highlight ability to capture significant cracks’ orientation, length, and growth stages, essential for understanding fatigue mechanisms. Analysis of results, especially evaluation metrics encompassing false positives and false negatives, provides a comprehensive understanding of the algorithm’s strengths and limitations. The proposed tool is available on GitHub.
An improved methodology for accelerated marine immersed corrosion testing of ship structural components
- Krzysztof Wołoszyk
- Emil Roch
- Beata Zima
- Yordan Garbatov
2024 Ships and Offshore Structures
Corrosion degradation is a prominent ageing mechanism in engineering structures (e.g. ships and offshore structures), leading to safety concerns and significant economic expenses. This study presents an improved methodology for indoor accelerated corrosion testing of structural components. The method involves controlling natural factors (oxygen saturation, temperature, salinity, and flow rate) to accelerate the corrosion in a controlled environment. The study uses small coupons, medium specimens, and stiffened plates made of steel with different thicknesses. Significant degradation acceleration was achieved compared to natural conditions (approx. twenty times faster). The mean corrosion equals 3.3 mm/year with a maximum 10% variation for different specimen sizes and thicknesses. A novel approach to account for the mass of corrosion products during periodical measurements was proposed and validated. The improved methodology offers an efficient and accurate way to simulate marine immersed corrosion, enabling further research on corrosion degradation behaviour and resistance in ship structures.
An improvement of body surface area formulas using the 3D scanning technique
- Grzegorz Redlarski
- Sławomir Kozieł
- Marek Krawczuk
- Janusz Siebert
- Marek Tałałaj
- Aleksander Pałkowski
- Piotr Tojza
- Zuzanna Cieślikowska
- Leszek Litzbarski
2024 Pełny tekst International Journal of Occupational Medicine and Environmental Health
Objectives: Body surface area (BSA) is one of the major parameters used in several medical fields. However, there are concerns raised about its usefulness, mostly due to the ambiguity of its estimation. Material and Methods: Authors have conducted a voluntary study to investigate BSA distribution and estimation in a group of 179 adult people of various sex, age, and physique. Here, there is provided an extended analysis of the majority of known BSA formulas. Furthermore, it was supplement with a comparison with the authors' propositions of enhanced formulas coefficients for known formulas models as well as with new power models based on an increased number of anthropometric data. Results: Introduction of the enhanced formulas coefficients cause a reduction of at least 30.5% in mean absolute error and 21.1% in maximum error in comparison with their known counterparts. Conclusions: In the context of the analysis presented it can be stated that the development of a single universal body surface area formula, based on a small number of state variables, is not possible. Therefore, it is necessary and justified to search for new estimation models that allow for quick and accurate calculation of body surface area for the entire population, regardless of individual body variations. The new formulas presented are such an alternative, which achieves better results than the previously known methods. Int J Occup Med Environ Health 2024;37(2)
An Innovative Floating System with a Savonius Rotor as a Horizontal-Axis Wind Turbine
- Joanna Grzelak
- Lara Guijarro Carrillo
- Jacek Nakielski
- Michał Piotrowicz
- Krzysztof Doerffer
2024 Pełny tekst Polish Maritime Research
In this project, an innovative wind turbine was designed for a floating plant. A large Savonius rotor was replaced with a double-rotor wind turbine implemented as a horizontal-axis turbine. This double rotor was positioned on the tip of a thrust plate and fixed to the deck of a catamaran. Simple 2D numerical simulations were performed to confirm the effectiveness of the concept. An analysis of the floating system configuration was carried out, and the loads and stresses on the system components were verified. Next, floating supports with appropriate sizes were selected to counteract the forces on the wind turbine system. Finally, an anchoring system with full rotational freedom was selected for the f loating platform. The present work was conducted as part of a Master’s thesis.
An Innovative New Approach to Light Pollution Measurement by Drone
- Katarzyna Bobkowska
- Paweł Burdziakowski
- Paweł Tysiąc
- Mariusz Pulas
2024 Drones
The study of light pollution is a relatively new and specific field of measurement. The current literature is dominated by articles that describe the use of ground and satellite data as a source of information on light pollution. However, there is a need to study the phenomenon on a microscale, i.e., locally within small locations such as housing estates, parks, buildings, or even inside buildings. Therefore, there is an important need to measure light pollution at a lower level, at the low level of the skyline. In this paper, the authors present a new drone design for light pollution measurement. A completely new original design for an unmanned platform for light pollution measurement is presented, which is adapted to mount custom sensors (not originally designed to be mounted on a unmanned aerial vehicles) allowing registration in the nadir and zenith directions. The application and use of traditional photometric sensors in the new configuration, such as the spectrometer and the sky quality meter (SQM), is presented. A multispectral camera for nighttime measurements, a calibrated visible-light camera, is used. The results of the unmanned aerial vehicle (UAV) are generated products that allow the visualisation of multimodal photometric data together with the presence of a geographic coordinate system. This paper also presents the results from field experiments during which the light spectrum is measured with the installed sensors. As the results show, measurements at night, especially with multispectral cameras, allow the assessment of the spectrum emitted by street lamps, while the measurement of the sky quality depends on the flight height only up to a 10 m above ground level.
An intelligent cellular automaton scheme for modelling forest fires
- Joan Boters Pitarch
- María Signes-Pont
- Julian Szymański
- Higinio Mora-Mora
2024 Pełny tekst Ecological Informatics
Forest fires have devastating consequences for the environment, the economy and human lives. Understanding their dynamics is therefore crucial for planning the resources allocated to combat them effectively. In a world where the incidence of such phenomena is increasing every year, the demand for efficient and accurate computational models is becoming increasingly necessary. In this study, we perform a revision of an initial proposal which consists of a two-dimensional propagation model based on cellular automata (2D-CA), which aims to understand the dynamics of these phenomena. We identify the key theoretical weaknesses and propose improvements to address these limitations. We also assess the effectiveness and accuracy of the model by evaluating improvements using real forest fire data (Beneixama, Alicante 2019). Moreover, as a result of the theoretical modifications performed, we introduce a novel intelligent architecture that seeks to capture relationships between system cells from the data. This new architecture has the ability to advance our understanding of forest fire dynamics, contributing to both the evaluation of existing protocols and more efficient firefighting resource management.
An inverse algorithm for contact heat conduction problems with an interfacial heat source based on a first-order thermocouple model
- Oleksii Nosko
2024 Pełny tekst INTERNATIONAL COMMUNICATIONS IN HEAT AND MASS TRANSFER
Inverse problems of contact heat conduction with an interfacial heat source are common in various fields of science, engineering and technology. In this study, an algorithm for their solution is developed based on an inverse parametric optimisation method with an impulse response function describing the heat partition and contact heat transfer. A first-order thermocouple model with a time constant parameter is embedded in the impulse response function. The specific power of the heat source is sought in the form of a polynomial from the condition of least-squares deviation of the simulated temperature from the temperature samples obtained by a thermocouple. Compared to the classical methods of simple inverse convolution and sequential function specification, the algorithm proves to be accurate in a substantially larger region of variation of the heating duration and time constant, covering slow-response thermocouple measurements. Additionally, the algorithm is significantly more robust against noise with a sufficient number of temperature samples. The applicability of the algorithm is demonstrated by solving inverse problems of contact heat conduction typical for sliding friction, laser and electric resistance welding at different thermal contact conditions and ratios of the time constant to the heating duration.
An objective isogeometric mixed finite element formulation for nonlinear elastodynamic beams with incompatible warping strains
- Myung-Jin Choi
- Sven Klinkel
- S. Klarmann
- Roger Sauer
2024 MULTIBODY SYSTEM DYNAMICS
We present a stable mixed isogeometric finite element formulation for geometrically and materially nonlinear beams in transient elastodynamics, where a Cosserat beam formulation with extensible directors is used. The extensible directors yield a linear configuration space incorporating constant in-plane cross-sectional strains. Higher-order (incompatible) strains are introduced to correct stiffness, whose additional degrees of freedom are eliminated by an element-wise condensation. Further, the present discretization of the initial director field leads to the objectivity of approximated strain measures, regardless of the degree of basis functions. For physical stress resultants and strains, we employ a global patch-wise approximation using B-spline basis functions, whose higher-order continuity enables using much fewer degrees of freedom than an element-wise approximation. For time-stepping, we employ implicit energy–momentum consistent scheme, which exhibits superior numerical stability in comparison to standard trapezoidal and mid-point rules. Several numerical examples are presented to verify the present method.
An Open Platform Tool for 2D Multipactor Simulations in Metallic Microwave Components
- Tomasz Nałęcz
- Łukasz Nowicki
- Małgorzata Celuch
- Michał Baranowski
- Adam Lamęcki
- Michał Mrozowski
2024 Pełny tekst
The paper presents a computer simulation software aimed at assessing the multipactor threshold power in a rectangular waveguide working with single tone excitation. Initial tests demonstrate a strong agreement between the simulation results obtained and those from commercial software. Contrary to the existing commercial software, our tool will be provided as Open Platform, for free use and popularisation of knowledge about physical phenomena resulting from interactions of microwaves with materials.
An optimized dissolved oxygen concentration control in SBR with the use of adaptive and predictive control schemes
- Tomasz Zubowicz
- Tomasz Ujazdowski
- Zuzanna Klawikowska
- Robert Piotrowski
2024 Pełny tekst
This paper addresses the problem of optimizing control of the aeration process in a water resource recovery facility (WRRF) using sequencing batch reactor (SBR), one that affects the efficiency of wastewater treatment by stimulating metabolic reactions of microorganisms through dissolved oxygen (DO) level control, and accounts for the predominant part of operating costs. Two independent approaches to DO control algorithm design based on nonlinear model-based predictive control (NMPC) with constraints and direct model reference adaptive control (DMRAC) are proposed and compared. Both algorithms were developed on the basis of utility models obtained by cognitive model simplification, however, both algorithms are characterized by a distinct mechanism to achieve control optimality and incorporate uncertainty. The NMPCbased algorithm solves an online optimization task by reducing the impact of uncertainty through feedback and estimating its influence by evaluating the differences between the internal model and measurements on a sliding prediction window. In contrast, DMRAC reduces the impact of uncertainty through the adaptation of control law parameters. Meanwhile, optimality is encoded in the reference model parameters reflecting the operation of the closed-loop system and in the independent parameters of the adaptation mechanism. Illustrations of the algorithms’ operation were provided by simulation experiments using a three-layer SBR model of the Swarzewo wastewater treatment plant with ASM3e-based reactions.
An Optimized Ka-Band Low Profile Dual-Polarized Transmitarray Antenna With 2D Beam Switching
- Seyed Hashem Ramazannia Tuloti
- Adam Lamęcki
- Michał Mrozowski
2024 Pełny tekst IEEE Access
This article presents an optimized dual-polarized transmitarray antenna (TA) designed for MIMO applications at the Ka-band, capable of switching beams in two directions. The antenna aperture uses a small unit cell with three layers of Taconic RF-35 dielectric substrates, which can be easily fabricated using PCB technology. The unit cell achieved a 360-degree phase shift and a transmission magnitude exceeding –0.4 dB at 28 GHz. We used nine dual-polarized patch antennas in a cross shape, each with a 10.5 dBi gain at 28 GHz, to switch the beams in two directions without changing the feed location. We optimized the phase distribution in the TA aperture and adjusted the feed antenna’s F/D and tilt to achieve a high-gain antenna with low-gain roll-off during beam switching. The fabricated TA exhibited excellent agreement with the full-wave simulation results. It achieved ±15 degrees and ±30 degrees beam tilts in the x- and y- directions, with less than 0.8 dB gain roll-off for both polarizations