2023 - Repozytorium publikacji - Politechnika Gdańska

Pokaż publikacje z roku

Publikacje z roku 2023

Pokaż wszystkie

Lipopolysaccharides: regulated biosynthesis and structural diversity
- Satish Raina
2023 Pełny tekst INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
The cell envelope of Gram-negative bacteria contains two distinct membranes, an inner (IM) and an outer (OM) membrane, separated by the periplasm, a hydrophilic compartment that includes a thin layer of peptidoglycan. The most distinguishing feature of such bacteria is the presence of an asymmetric OM with phospholipids located in the inner leaflet and lipopolysaccharides (LPSs) facing the outer leaflet. The maintenance of this OM asymmetry is essential to impart a permeability barrier, which prevents the entry of bulky toxic molecules, such as antibiotics and bile salts, into the cells [1]. LPS is a complex glycolipid that, with few exceptions, is essential for bacterial viability and is one of the major virulence factors in pathogenic Gram-negative bacteria. Model bacteria, such as Escherichia coli, contain approximately 2–3 × 106 molecules of LPS that cover more than 75% of the OM [2]. The composition of LPS is highly heterogenous; however, they often share a common architecture, and for convenience can be divided into three parts. The first, a conserved glycophospholipid moiety called lipid A, which anchors LPS in the OM, constitutes the endotoxin principal since it is recognized by the innate immune cell receptor TLR4/MD2-CD14 complex. A proximal core oligosaccharide is attached to lipid A via 3-deoxy-α-D-manno-oct-2-ulsonic acid (Kdo), and in smooth bacteria a distal O-polysaccharide called an O-antigen is attached [3]. It should be noted that some bacteria display LPSs without the O-chain, which are thus named lipooligosaccharides (LOSs). The biosynthesis of LPS begins with the formation of the essential key precursor molecule Kdo2-lipid A, which requires the sequential action of seven essential enzymes on the cytoplasmic side of the IM. This Kdo2-lipid A serves as a substrate for an extension by the incorporation of various sugars by specific glycosyltransferases before the lipid A-core molecules are flipped by MsbA to the periplasmic side of the IM.
Load introduction to composite columns revisited—Significance of force allocation and shear connection stiffness
- B. Grzeszykowski
- Maciej Lewandowski-Szewczyk
- M. Niedośpiał
2023 Pełny tekst ENGINEERING STRUCTURES
The AISC 360-16 Specification recommends that the design shear force between parts of a composite column in the load introduction area shall be calculated based on the force allocation at ultimate limit state. Applicability of this straightforward method to the load levels that usually arise in slender composite columns is questionable, as this capacity-based force allocation is only true when the axial force is equal to the plastic resistance of the composite cross-section. Next, the number of required shear connectors is calculated as a quotient of the design shear force and the strength of a single shear connector. We demonstrate that: first, for the lower load levels, the stiffness-based force allocation gives a more accurate estimate of the shear force; second, the number of shear connectors satisfying the strength requirement can lead to insufficient force transfer between parts of the composite cross-section. To investigate the shear transfer mechanism in composite columns, we derive an analytical model with linear elastic constitutive relations both for steel and concrete and three types of shear force slip laws: elastic, elastic plastic, and rigid plastic. The case studies carried out for different shear transfer scenarios demonstrate the importance of the shear connection stiffness on the effectiveness of the load introduction. The remaining portion of the shear force is transferred outside the load introduction area, which hampers the column's ability to withstand shearing from varying bending moments or incipient buckling. To control the shear force transfer efficiency by enhancing the shear connection stiffness, we propose an original Stiffness Method and provide design charts as an aid in the design process.
Longitudinal drug synergy assessment using convolutional neural network image-decoding of glioblastoma single-spheroid cultures
- Anna Giczewska
- Krzysztof Pastuszak
- Megan Houweling
- U Kulsoom Abdul
- Noa Faaij
- Laurine Wedekind
- David Noske
- Thomas Würdinger
- Anna Supernat
- Bart Westerman
2023 Pełny tekst Neuro-Oncology Advances
Abstract Background In recent years, drug combinations have become increasingly popular to improve therapeutic outcomes in various diseases, including difficult to cure cancers such as the brain cancer glioblastoma. Assessing the interaction between drugs over time is critical for predicting drug combination effectiveness and minimizing the risk of therapy resistance. However, as viability readouts of drug combination experiments are commonly performed as an endpoint where cells are lysed, longitudinal drug-interaction monitoring is currently only possible through combined endpoint assays. Methods We provide a method for massive parallel monitoring of drug interactions for 16 drug combinations in three glioblastoma models over a time frame of 18 days. In our assay, viabilities of single neurospheres are to be estimated based on image information taken at different time points. Neurosphere images taken at the final day (day 18) were matched to the respective viability measured by CellTiter-Glo 3D at the same day. This allowed to use machine learning to decode image information to viability values at day 18 as well as for the earlier time points (at day 8, 11, 15). Results Our study shows that neurosphere images allow to predict cell viability from extrapolated viabilities. This enables to assess the drug interactions in a time-window of 18 days. Our results show a clear and persistent synergistic interaction for several drug combinations over time. Conclusions Our method facilitates longitudinal drug-interaction assessment, providing new insights into the temporal-dynamic effects of drug combinations in 3D neurospheres which can help to identify more effective therapies against glioblastoma.
Long-term mortality after transcatheter aortic valve implantation for aortic stenosis in immunosuppression-treated patients: a propensity-matched multicentre retrospective registry-based analysis
- Michał Walczewski
- Aleksandra Gąsecka
- Adam Witkowski
- Maciej Dabrowski
- Zenon Huczek
- Radosław Wilimski
- Andrzej Ochała
- Radosław Parma
- Bartosz Rymuza
- Marek Grygier
- Marek Jemielity
- Anna Olasińska-Wiśniewska
- Dariusz Jagielak
- Radosław Targoński
- Krzysztof Pastuszak
- Peter Gresner
- Marcin Grabowski
- Janusz Kochman
2023 Pełny tekst Postępy w Kardiologii Interwencyjnej
Introduction Data regarding patients with a previous medical record of immunosuppression treatment who have undergone transcatheter aortic valve implantation (TAVI) are limited and extremely inconclusive. Available studies are mostly short term observations; thus there is a lack of evidence on efficacy and safety of TAVI in this specific group of patients. Aim To compare the in-hospital and long-term outcomes between patients with or without a medical history of immunosuppressive treatment undergoing TAVI for aortic valve stenosis (AS). Material and methods We conducted a retrospective registry-based analysis including patients undergoing TAVI for AS at 5 centres between January 2009 and August 2017. The primary endpoint was long-term all-cause mortality. Secondary endpoints comprised major vascular complications, life-threatening or disabling bleeding, stroke and new pacemaker implantation. Results Of 1451 consecutive patients who underwent TAVI, two propensity-matched groups including 25 patients with a history of immunosuppression and 75 patients without it were analysed. No differences between groups in all-cause mortality were found in a median follow-up time of 2.7 years following TAVI (p = 0.465; HR = 0.73; 95% CI: 0.30–1.77). The rate of major vascular complications (4.0% vs. 5.3%) was similar in the two groups (p = 1.000). There were no statistically significant differences in the composite endpoint combining life-threatening or disabling bleeding, major vascular complications, stroke and new pacemaker implantation (40.0% vs. 20.0%, p = 0.218). Conclusions Patients who had undergone TAVI for AS had similar long-term mortality regardless of whether they had a previous medical record of immunosuppression. Procedural complication rates were comparable between the groups.
Long‐time scale simulations of virus‐like particles from three human‐norovirus strains
- Agnieszka Lipska
- Adam Sieradzan
- Cezary Czaplewski
- Andrea D. Lipińska
- Krzysztof Ocetkiewicz
- Jerzy Proficz
- Paweł Czarnul
- Henryk Krawczyk
- Józef Liwo
2023 Pełny tekst JOURNAL OF COMPUTATIONAL CHEMISTRY
The dynamics of the virus like particles (VLPs) corresponding to the GII.4 Houston, GII.2 SMV, and GI.1 Norwalk strains of human noroviruses (HuNoV) that cause gastroenteritis was investigated by means of long-time (about 30 μs in the laboratory timescale) molecular dynamics simulations with the coarse-grained UNRES force field. The main motion of VLP units turned out to be the bending at the junction between the P1 subdomain (that sits in the VLP shell) and the P2 subdomain (that protrudes outside) of the major VP1 protein, this resulting in a correlated wagging motion of the P2 subdomains with respect to the VLP surface. The fluctuations of the P2 subdomain were found to be more pronounced and the P2 domain made a greater angle with the normal to the VLP surface for the GII.2 strain, which could explain the inability of this strain to bind the histo-blood group antigens (HBGAs).
LOS and NLOS identification in real indoor environment using deep learning approach
- Alicja Olejniczak
- Olga Błaszkiewicz
- Krzysztof Cwalina
- Piotr Rajchowski
- Jarosław Sadowski
2023 Pełny tekst Digital Communications and Networks
Visibility conditions between antennas, i.e. Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) can be crucial in the context of indoor localization, for which detecting the NLOS condition and further correcting constant position estimation errors or allocating resources can reduce the negative influence of multipath propagation on wireless communication and positioning. In this paper a deep learning (DL) model to classify LOS/NLOS condition while analyzing two Channel Impulse Response (CIR) parameters: Total Power (TP) [dBm] and First Path Power (FP) [dBm] is proposed. The experiments were conducted using DWM1000 DecaWave radio module based on measurements collected in a real indoor environment and the proposed architecture provides LOS/NLOS identification with an accuracy of more than 100% and 95% in static and dynamic senarios, respectively. The proposed model improves the classification rate by 2-5% compared to other machine learning (ML) methods proposed in the literature.
(Lost) Pride and Prejudice. Journalistic Identity Negotiation Versus the Automation of Content
- Jan Kreft
- Monika Boguszewicz-Kreft
- Mariana Fydrych
2023 Journalism Practice
The objective of our research was to broaden the knowledge regarding the relationship between the work of journalists and their professional identity, and, in particular, to identify the attitudes of this professional group towards algorithmic content creation under conditions of liminality. Previously, the implementation of the technology of algorithmic content creation by media organisations was associated primarily with financial factors (production savings). The pandemic situation, for security reasons forcing the use of new technologies to perform remote work, became an additional factor enhancing the sense of liminality. A qualitative study was conducted in the form of 25 in-depth interviews in leading Polish media at the initial stage of the pandemic. The results showed that the most important aspect concerning liminality was the loss of pride in performing a prestigious profession. Following waves of financial savings in editorial offices, and after the pandemic, journalists viewed the algorithmic creation of content as the next potential “plague” affecting their perceived degradation of the profession. The anticipated change in working conditions, already interpreted as a threat to journalists, signified a liminal experience dictated by a new factor and prompted them to choose defence strategies.
Low-cost 3D Printed Circularly Polarized Lens Antenna for 5.9 GHz V2X Applications
- Weronika Kalista
- Luiza Leszkowska
- Mateusz Rzymowski
- Krzysztof Nyka
- Łukasz Kulas
2023
This paper presents design and realization of a circularly polarized antenna consisting of a linearly polarized patch antenna and a 3D printed lens, at the same time performing the functions of wave collimator and a polarizer. The antenna is dedicated for 802.11p systems, as a part of road infrastructure, with operation bandwidth 5.85 - 5.925 GHz. Its realised gain and axial ratio at center frequency 5.9 GHz are 14.3 dBi and 2.17 dB respectively. The lens provides approximately 6% bandwidth with axial ratio below 3 dB. The proposed antenna is easy to design and fabricate and can be realized with the use of low-cost materials.
Low-Cost and Highly-Accurate Behavioral Modeling of Antenna Structures by Means of Knowledge-Based Domain-Constrained Deep Learning Surrogates
- Sławomir Kozieł
- Nurullah Calik
- Peyman Mahouti
- Mehmet Belen
2023 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION
The awareness and practical benefits of behavioral modeling methods have been steadily growing in the antenna engineering community over the last decade or so. Undoubtedly, the most important advantage thereof is a possibility of a dramatic reduction of computational expenses associated with computer-aided design procedures, especially those relying on full-wave electromagnetic (EM) simulations. In particular, the employment of fast replacement models (surrogates) allows for repetitive evaluations of the antenna structure at negligible cost, thereby accelerating processes such as parametric optimization, multi-criterial design, or uncertainty quantification. Notwithstanding, a construction of reliable data-driven surrogates is seriously hindered by the curse of dimensionality and the need for covering broad ranges of geometry/material parameters, which is imperative from the perspective of design utility. A recently proposed constrained modeling approach with knowledge-based stochastic determination of the model domain addresses this issue to a large extent and has been demonstrated to enable quasi-global modeling capability while maintaining a low setup cost. This work introduces a novel technique that capitalizes on the domain confinement paradigm and incorporates deep-learning-based regression modeling to facilitate handling of highly-nonlinear antenna characteristics. The presented framework is demonstrated using three microstrip antennas and favorably compared to several state-of-the-art techniques. The predictive power of our models reaches remarkable two percent of a relative RMS error (averaged over the considered antenna structures), which is a significant improvement over all benchmark methods.
Low-Cost Behavioral Modeling of Antennas by Dimensionality Reduction and Domain Confinement
- Sławomir Kozieł
- Anna Pietrenko-Dąbrowska
- Leifur Leifsson
2023
Behavioral modeling has been rising in importance in modern antenna design. It is primarily employed to diminish the computational cost of procedures involving massive full-wave electromagnetic (EM) simulations. Cheaper alternative offer surrogate models, yet, setting up data-driven surrogates is impeded by, among others, the curse of dimensionality. This article introduces a novel approach to reduced-cost surrogate modeling of antenna structures, which focuses the modeling process on design space regions containing high-quality designs, identified by randomized pre-screening. A supplementary dimensionality reduction is applied via the spectral analysis of the random observable set. The reduction process identifies the most important directions from the standpoint of geometry parameter correlations, and spans the domain along a small subset thereof. As demonstrated, domain confinement as outlined above permits a dramatic improvement of surrogate accuracy in comparison to the state-of-the-art modeling approaches.
Low-Cost Open-Hardware System for Measurements of Antenna Far-Field Characteristics in Non-Anechoic Environments
- Jan Olencki
- Vorya Waladi
- Adrian Bekasiewicz
2023 Pełny tekst
Experimental validation belongs to the most important steps in the development of antenna structures. Measurements are normally performed in expensive, dedicated facilities such as anechoic chambers, or open-test sites. A high cost of their construction might not be justified when the main goal of antenna verification boils down to demonstration of the measurement procedure, or rough validation of the simulation models used for the development of the structure. Although solutions for far-field measurement of antennas in non-anechoic environments have been demonstrated in the literature, they utilize expensive equipment. In this work, a low-cost (around 3300 USD), system for experimental validation of antenna prototypes in non-anechoic conditions has been discussed. Its main components include the in-house developed heads and an open-hardware-based vector network analyzer. Performance of the system has been demonstrated using two antenna structures for which radiation patterns have been obtained. Comparisons against measurements performed in the anechoic chamber and using other expensive equipment have also been provided.
Low-frequency noise in Au-decorated graphene–Si Schottky barrier diode at selected ambient gases
- Janusz Smulko
- Katarzyna Drozdowska
- Adil Rehman
- Tesfalem Welearegay
- Lars Österlund
- Sergey Rumyantsev
- Grzegorz Cywiński
- Bartłomiej Stonio
- Aleksandra Krajewska
- Maciej Filipiak
- Pavlo Sai
2023 Pełny tekst APPLIED PHYSICS LETTERS
We report results of the current–voltage characteristics and low-frequency noise in Au nanoparticle (AuNP)-decorated graphene–Si Schottky barrier diodes. Measurements were conducted in ambient air with addition of either of two organic vapors, tetrahydrofuran [(CH2)4O; THF] and chloroform (CHCl3), as also during yellow light illumination (592nm), close to the measured particle plasmon polariton frequency of the Au nanoparticle layer. We observed a shift of the DC characteristics at forward voltages (forward resistance region) when tetrahydrofuran vapor was admitted (in a Au-decorated graphene–Si Schottky diode), and a tiny shift under yellow irradiation when chloroform was added (in not decorated graphene–Si Schottky diode). Signiﬁcantly larger difference in the low-frequency noise was observed for the two gases during yellow light irradiation, compared with no illumination. The noise intensity was suppressed by AuNPs when compared with noise in graphene–Si Schottky diode without an AuNP layer. We conclude that ﬂicker noise generated in the investigated Audecorated Schottky diodes can be utilized for gas detection.
Low-frequency noise in ZrS3 van der Waals semiconductor nanoribbons
- Adil Rehman
- Grzegorz Cywiński
- W. Knap
- Janusz Smulko
- Alexander Balandin
- Sergey Rumyantsev
2023 Pełny tekst APPLIED PHYSICS LETTERS
We report the results of the investigation of low-frequency electronic noise in ZrS3 van der Waals semiconductor nanoribbons. The test structures were of the back-gated field-effect-transistor type with a normally off n-channel and an on-to-off ratio of up to four orders of magnitude. The current–voltage transfer characteristics revealed significant hysteresis owing to the presence of deep levels. The noise in ZrS3 nanoribbons had spectral density SI ~ 1/f^c (f is the frequency) with c ~ 1.3–1.4 within the whole range of the drain and gate bias voltages. We used light illumination to establish that the noise is due to generation–recombination, owing to the presence of deep levels, and determined the energies of the defects that act as the carrier trapping centers in ZrS3 nanoribbons.
Low-Voltage LDO Regulator Based on Native MOS Transistor with Improved PSR and Fast Response
- Grzegorz Blakiewicz
2023 Pełny tekst ENERGIES
In this paper, a low-voltage low-dropout analog regulator (ALDO) based on a native n-channel MOS transistor is proposed. Application of the native transistor with the threshold voltage close to zero allows elimination of the charge pump in low-voltage regulators using the pass element in a common drain configuration. Such a native pass transistor configuration allows simplification of regulator design and improved performance, with supply voltages below 1 V, compared to commonly used regulators with p-channel MOS transistors. In the presented design of ALDO regulator in 180 nm CMOS X-FAB technology, an output voltage of 0.7 V was achieved with an output current of 10 mA and a supply voltage of 0.8 V. Simulation results show that despite the low supply voltage, output voltage spikes do not exceed 70 mV at the worst technology corner when output current transients from 100 uA to 10 mA. Under such conditions, stable operation and power supply rejection PSR = 35 dB were achieved with an output capacitance of 0–500 pF. The proposed regulator allows to push the limit of ALDO regulator applications to voltages below 1 V with only slight degradation of its performance.
Low-volume label-free SARS-CoV-2 detection with the microcavity-based optical fiber sensor
- Monika Janik
- Tomasz Gabler
- Marcin Koba
- Mirosława Panasiuk
- Yanina Dashkevich
- Tomasz Łęga
- Agnieszka Dąbrowska
- Antonina Naskalska
- Sabina Żołędowska
- Dawid Nidzworski
- Krzysztof Pyrć
- Beata Gromadzka
- Mateusz Śmietana
2023 Pełny tekst Scientific Reports
Accurate and fast detection of viruses is crucial for controlling outbreaks of many diseases; therefore, to date, numerous sensing systems for their detection have been studied. On top of the performance of these sensing systems, the availability of biorecognition elements specific to especially the new etiological agents is an additional fundamental challenge. Therefore, besides high sensitivity and selectivity, such advantages as the size of the sensor and possibly low volume of analyzed samples are also important, especially at the stage of evaluating the receptor-target interactions in the case of new etiological agents when typically, only tiny amounts of the receptor are available for testing. This work introduces a real-time, highly miniaturized sensing solution based on microcavity in-line Mach–Zehnder interferometer (μIMZI) induced in optical fiber for SARS-CoV-2 virus-like particles detection. The assay is designed to detect conserved regions of the SARS-CoV-2 viral particles in a sample with a volume as small as hundreds of picoliters, reaching the detection limit at the single ng per mL level.
Machine Learning Assisted Interactive Multi-objectives Optimization Framework: A Proposed Formulation and Method for Overtime Planning in Software Development Projects
- Hammed Mojeed
- Rafał Szłapczyński
2023
Machine Learning Assisted Interactive Multi-objectives Optimization Framework: A Proposed Formulation and Method for Overtime Planning in Software Development Projects Hammed A. Mojeed & Rafal Szlapczynski Conference paper First Online: 14 September 2023 161 Accesses Part of the Lecture Notes in Computer Science book series (LNAI,volume 14125) Abstract Software development project requires proper planning to mitigate risk and uncertainty. Overtime planning within software project management has been receiving attention recently from search-based software engineering researchers. Multi-objective evolutionary algorithms are used to build automated tools that could effectively help Project Managers (PM) plan overtime on project schedules. Existing models however lack applicability by the PMs due to their disregard for expert knowledge in planning overtime. This study proposes a new interactive problem formulation for software overtime planning and presents a framework for building a machine learning-based interactive multi-objective optimization algorithm for overtime planning in software development projects. The framework is designed to train a priori a machine learning model to mimic the PM’s subjective judgment of overtime plans within the project schedule. The machine learning model is integrated with a memetic multi-objective optimization algorithm via an interactive module. Also, the memetic algorithm incorporates a preference-based w-dominance method for selecting non-dominated solutions. The proposed framework will be developed to assist software project managers to better plan overtime in order to prevent the expected risk of software development overrun
Machine learning-based prediction of preplaced aggregate concrete characteristics
- Farzam Omidi Moaf
- Farzin Kazemi
- Hakim S. Abdelgader
- Marzena Kurpińska
2023 ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
Preplaced-Aggregate Concrete (PAC) is a type of preplaced concrete where coarse aggregate is placed in the mold and a Portland cement-sand grout with admixtures is injected to fill the voids. Due to the complex nature of PAC, many studies were conducted to determine the effects of admixtures and the compressive and tensile strengths of PAC. Considering that a prediction tool is needed to estimate the compressive and tensile strengths of PAC, this research developed 12 supervised Machine Learning (ML) algorithms in Python software to provide estimations for civil engineers. To prepare the training and testing datasets, a comprehensive investigation was performed to prepare experimental studies on the compressive and tensile strengths of PAC. Then, according to the features of the dataset, four scenarios were defined based on the input features. The capability of ML algorithms was investigated in each scenario. Results showed that the ETR, RDF, and BR algorithms achieved the prediction accuracy of 98.3%, 95.3% and 94.6%, respectively, for estimating the compressive strength of PAC with input features of Case B. Therefore, due to the performance of the ML models, their generality was investigated by preparing the experimental test of two specimens of PAC and by validating the results. Notably, that the proposed ML models (e.g. BR method) can accurately predict the compressive and tensile strengths of specimens (e.g. with accuracy of 98.4 99.7%, respectively) and can be used to facilitate and reduce the experimental tests as well as the experimental efforts.
Machine learning-based prediction of residual drift and seismic risk assessment of steel moment-resisting frames considering soil-structure interaction
- Neda Asgarkhani
- Farzin Kazemi
- Robert Jankowski
2023 COMPUTERS & STRUCTURES
Nowadays, due to improvements in seismic codes and computational devices, retrofitting buildings is an important topic, in which, permanent deformation of buildings, known as Residual Interstory Drift Ratio (RIDR), plays a crucial role. To provide an accurate yet reliable prediction model, 32 improved Machine Learning (ML) algorithms were considered using the Python software to investigate the best method for estimating Maximum Interstory Drift Ratio (IDRmax) and RIDR of 384 Steel Moment-Resisting Frames (SMRFs). In addition, the curve plot ability of methods was investigated to provide an estimation of Median of IDA curve (IDAMed) and Seismic Failure Probability curve (SFPCurve) considering Soil-Structure Interaction (SSI) effects. It is noteworthy that ML algorithms were improved with a pipeline-based hyper-parameters Fine-Tuning (FT) method followed by forward and backward feature selection methodologies to avoid overfitting and data leakage issues. The improved methods were evaluated to find the best prediction model regarding seismic demands. The results show that proposed methods have higher prediction accuracy and curve fitting ability (i.e. more than 95%) that can be used to estimate IDAMed and SFPCurve of a structure to accelerate the seismic risk assessment. A prediction tool is introduced to use the methods of this study for estimating abovementioned seismic demands.
Machine learning-based prediction of seismic limit-state capacity of steel moment-resisting frames considering soil-structure interaction
- Farzin Kazemi
- Robert Jankowski
2023 COMPUTERS & STRUCTURES
Regarding the unpredictable and complex nature of seismic excitations, there is a need for vulnerability assessment of newly constructed or existing structures. Predicting the seismic limit-state capacity of steel Moment-Resisting Frames (MRFs) can help designers to have a preliminary estimation and improve their views about the seismic performance of the designed structure. This study improved data-driven decision techniques in Python software, known as supervised Machine Learning (ML) algorithms, to find median IDA curves (M-IDAs) for predicting the seismic limit-state capacities of steel MRFs considering Soil-Structure Interaction (SSI) effects. For this purpose, Incremental Dynamic Analyses (IDAs) were per-formed on the steel MRFs from two to nine-story elevations modeled in Opensees subjected to three ground motion subsets of Far Fault (FF), near-fault Pulse-Like (PL) and No-Pulse (NP) suggested by FEMA-P695. The result of the analysis confirmed that there is no specific model for predicting the M-IDA curve of steel structures; therefore, the best developed ML algorithms to reduce a complex modelling process with high computational cost using 128,000 data points were proposed. To provide convenient access to prediction results, Graphical User Interface (GUI) was developed to predict Sa (T1) of seismic limit-state performance levels with a large database based on prediction models.
Machine learning-based seismic fragility and seismic vulnerability assessment of reinforced concrete structures
- Farzin Kazemi
- Neda Asgarkhani
- Robert Jankowski
2023 Pełny tekst SOIL DYNAMICS AND EARTHQUAKE ENGINEERING
Many studies have been performed to put quantifying uncertainties into the seismic risk assessment of reinforced concrete (RC) buildings. This paper provides a risk-assessment support tool for purpose of retrofitting and potential design strategies of RC buildings. Machine Learning (ML) algorithms were developed in Python software by innovative methods of hyperparameter optimization, such as halving search, grid search, random search, fine-tuning method, and the k-fold cross-validation, to derive the seismic fragility curve for accelerating seismic risk assessment. Proposed ML methods significantly reduced the computational efforts compared to conventional procedure of seismic fragility assessment. The prediction results can be combined with considered hazard curves for the purpose of seismic risk assessment of RC buildings. To prepare the training dataset, Incremental Dynamic Analyses (IDAs) were performed on 165 RC frames to achieve 1121184 data points. Performance indicators showed that the algorithms of Artificial Neural Networks (ANNs), Extra-Trees Regressor (ETR), Extremely Randomized Tree Regressor (ERTR), Bagging Regressor (BR), Extreme Gradient Boosting (XGBoost), and Histogram-based Gradient Boosting Regression (HGBR) had higher performance, which achieved acceptable accuracy and fitted to actual curves. In addition, Graphical User Interface (GUI) was introduced as a practical tool yet reliable for seismic risk assessment of RC buildings.