Orsolya Kegyes-Brassai | Széchenyi István University (original) (raw)

Conference Presentations by Orsolya Kegyes-Brassai

Geophysical research abstracts, Mar 9, 2020

Papers by Orsolya Kegyes-Brassai

<p>One of the critical research areas of engineering science is the assessment of e... more <p>One of the critical research areas of engineering science is the assessment of existing buildings' seismic vulnerability. Since the existing building stock is made up of structures that were built before design codes were developed, taking into account low or moderate design codes may make them vulnerable to an upcoming earthquake. To this end, rapid visual screening (RVS) methods can be used to identify building vulnerability before or after an earthquake. A number of RVS methods have been developed; however, it's critical to assess their accuracy in terms of their ability to reliably classify the state of building vulnerability. Therefore, this study offers an application of conventional RVS methods (FEMA P-154, RISK-UE Project, JBDPA), as well as a comparison of the results of the building safety level classification of 20 existing reinforced concrete buildings from Győr, Hungary. Pushover analysis was used as a detailed vulnerability assessment technique in addition to the RVS methods to evaluate one more reinforced concrete building. The findings of pushover analysis and RVS methods are contrasted to demonstrate how effectively RVS methods can be utilized to determine building vulnerability. Additionally, the findings of this study can be used to select an RVS method for commencing a pre-earthquake building assessment of the existing reinforced concrete building stock in the investigation area.</p> <p><strong>Keywords:</strong> earthquake; existing buildings; vulnerability assessment; rapid visual screening; machine learning; building damage state</p>

Applied sciences, Oct 23, 2023

DOAJ (DOAJ: Directory of Open Access Journals), Sep 1, 2022

Our team has compiled Seismotectonic hazard map of Hungary. One of the main contents of the map a... more Our team has compiled Seismotectonic hazard map of Hungary. One of the main contents of the map are Eurocode 8 categories whose production steps are described here in detail. In engineering, site response to earthquakes has been classified to national and international standards. Eurocode 8 standard is partly based on Vs30 that is the time averaged shear-wave velocity in the uppermost 30 m sediment. We have compiled 67 Vs30 measurements and collected 103 Vs30 values from PhD theses and industrial reports. The values could be divided to soil class A–D of Eurocode 8 which are defined by Vs30 thresholds. The special soil class E (hard rock beneath 5–20 m thick loose sediment) needed a deeper investigation. The Vs trend was plotted and plots with obvious knickpoint has been analysed further. In case of one knickpoint in Vs trend two-layered model was used. We were defined the thickness and the theoretical Vs30 of the upper and the lower strata. In case if the site fit to class E, original Eurocode 8 class have been overwritten. Other advantage of the extrapolation of Vs trend of the uppermost strata is to derive the theoretical Vs30 of the given geomorphological feature if its sediment would fill up the whole 30 m.In Hungary only the youngest and lowest level of alluvial and lacustrine features fall into the most critical class D. Therefore that features have been mapped. In case of the youngest sediment’s thickness was not exceeded 20 m in each places, that site would classified as „shallow D” which is not a Eurocode 8 soil class. This process could be done using the borehole database of Geomega Ltd. Classification of soil class E have derived using the same method: thousands of borehole data have been checked to delineate the margin of the categories around the rock outcrops. For soil classes A–D topographical slope – Vs30 relation has established. For Hungary, we recommend to use 0.3%, 3% and 11% as topographical slope barriers between soil classes D-C-B-A (in advance).Secondly, active faults were mapped using the methodology described by the European Facilities for Earthquake Hazard and Risk. Third, earthquake database was use to present area affected by frequent ground motions. We have divided the database to historical and to instrumental detections due to their differences in the accuracy and reliability of magnitude and epicentre location.Historically Komárom-Oroszlány-Balatonfő line was most affected by earthquakes. Our map revealed that in the Middle Hungarian Shear Zone consists of still active fault lines. Some spots are affected by densely located small earthquakes such as the neighbourhood of Zalaszengrót, Répcelak, Nagyigmánd, the DIósjenő fault, Heves, Csepel, Jászberény, Nagykanizsa, Nagyatád, Pincehely, Szabadszállás, Kecskemét, and Miskolc. In almost all cases the most critical soil class D can be found in the neighbourhood of mentioned sites, while class E appears only in some locations.The research project was supported by the National Research, Development and Innovation Office of Hungary (2018-1.2.1-NKP-2018-00007). Map can be downloaded among others and vector data can be requested at Geomega website (www.geomega.hu).

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Acta Geodaetica et Geophysica, 2015

Determination of shear-wave (S-wave) velocity profiles is the first step in seismic hazard assess... more Determination of shear-wave (S-wave) velocity profiles is the first step in seismic hazard assessment of a town, because the dynamic parameters of local soil types are vital for seismic response analysis of a specific area in order to determine the local soil effect in a case of a seismic event for seismic risk analysis. S-wave velocity profiles have been determined for many areas within Gy} or. Extensive use of historical boring logs allowed for correlations and reasonable extrapolation of soil performance throughout the area. This has led to a pattern of soil layer distributions and delineates several different soil zones for Gy} or. Keywords Local site effect Á Shear-wave velocity profile Á Dynamic properties of soils 1 Introduction Research in earthquake hazard mitigation has focused on evaluating possible damage scenarios for different magnitude events (Luco et al. 2007; Committee on National Earthquake Resilience 2011). Although seismic events are rare in many places, they are characterized by high exposure and their economic and social effects cannot be neglected.

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

<p>One of the critical research areas of engineering science is the assessment of e... more <p>One of the critical research areas of engineering science is the assessment of existing buildings' seismic vulnerability. Since the existing building stock is made up of structures that were built before design codes were developed, taking into account low or moderate design codes may make them vulnerable to an upcoming earthquake. To this end, rapid visual screening (RVS) methods can be used to identify building vulnerability before or after an earthquake. A number of RVS methods have been developed; however, it's critical to assess their accuracy in terms of their ability to reliably classify the state of building vulnerability. Therefore, this study offers an application of conventional RVS methods (FEMA P-154, RISK-UE Project, JBDPA), as well as a comparison of the results of the building safety level classification of 20 existing reinforced concrete buildings from Győr, Hungary. Pushover analysis was used as a detailed vulnerability assessment technique in addition to the RVS methods to evaluate one more reinforced concrete building. The findings of pushover analysis and RVS methods are contrasted to demonstrate how effectively RVS methods can be utilized to determine building vulnerability. Additionally, the findings of this study can be used to select an RVS method for commencing a pre-earthquake building assessment of the existing reinforced concrete building stock in the investigation area.</p> <p><strong>Keywords:</strong> earthquake; existing buildings; vulnerability assessment; rapid visual screening; machine learning; building damage state</p>

Sustainability, Dec 6, 2022

This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY

Geosciences

In order to prevent possible loss of life and property, existing building stocks need to be asses... more In order to prevent possible loss of life and property, existing building stocks need to be assessed before an impending earthquake. Beyond the examination of large building stocks, rapid evaluation methods are required because the evaluation of even one building utilizing detailed vulnerability assessment methods is computationally expensive. Rapid visual screening (RVS) methods are used to screen and classify existing buildings in large building stocks in earthquake-prone zones prior to or after a catastrophic earthquake. Buildings are assessed using RVS procedures that take into consideration the distinctive features (such as irregularity, construction year, construction quality, and soil type) of each building, which each need to be considered separately. Substantially, studies have been presented to enhance conventional RVS methods in terms of truly identifying building safety levels by using computer algorithms (such as machine learning, fuzzy logic, and neural networks). This...

Earthquake Resistant Engineering Structures X, 2015

Assessment and management of earthquake risk requires several disciplines and different aspects t... more Assessment and management of earthquake risk requires several disciplines and different aspects to evaluate. Based on mathematical calculations, risk is the product of hazard and vulnerability. Considering earthquakes, the process of risk evaluation consists of several steps performed in parallel. One line of the research deals with the seismicity of the area by identifying potential sources, usually along fault systems, for probabilistic hazard assessment. The next step focuses on the simulation of strong ground motions for regional hazard assessment. Ground motion attenuation models are calibrated based on observed strong motion recordings and on theoretical computations taking into account realistic models for wave propagation as well as possible seismic source zones. Further research takes into account local site effects and will result in a microzonation for the evaluated area based on local soil conditions and field testing of the soil properties. Another line of research focuses on the built environment. This paper presents the case study for the city of Győr, focusing on the seismic hazard assessment of the typical soil types along the Danube. Extensive use of historical boring logs allowed for correlations and reasonable extrapolation of soil performance throughout the area. This has led to a pattern of soil layer distributions and delineates of several different soil zones for Győr. Compared to more simplified profile standards from building codes, it can be clearly seen that the different levels of hazard concerning Győr are apparent, even though almost all of the region belongs to soil category C based on EC8.

<p>Being part of the expert team sent by Hungarian National Directorate General for... more <p>Being part of the expert team sent by Hungarian National Directorate General for Disaster Management to Tirana after the 6.4 magnitude earthquake in 2019, and experienced the fear of residents living in slightly or heavily cracked buildings raised the question of preparedness to against seismic events in other moderate seismic regions such as Hungary. Recent earthquakes within the moderate range have proved that moderate seismicity does not necessarily equate to moderate damage suffered.</p><p>Vulnerability to earthquakes has increased due to extending urban areas. This paper presents lessons learnt after medium-sized earthquakes and examines the adaptability of measures taken afterwards and the possibility to apply these methods to other regions in Hungary and throughout Europe where the seismic hazard is not great, but cannot be ignored. To reduce the potential damage, a comprehensive assessment of the seismic risk followed by a package of relevant remedial measures is needed. Methods developed for Hungary is presented compared to methods applied in other regions to determine local site effects, vulnerability, and preparedness, being the main components responsible to risks. Based on the results, engineers can better plan to make improvements to infrastructure, and authorities can better plan for emergency activities in case of a seismic event.</p>

International Journal of GEOMATE, 2018

Closed-space fires often occur in Hungary, so it is necessary to examine the effects of fires on ... more Closed-space fires often occur in Hungary, so it is necessary to examine the effects of fires on building structures, taking into account Hungarian architectural features. Fires inside the buildings are characterized by intense heat development and smoke generation that can cause permanent damage to the building structures. Heat and smoke extraction during fire extinguishing is based usually on natural ventilation. Not only being a non-effective process also makes it more difficult to accomplish firefighting tasks. Experiments in this research have been conducted with mobile positive pressure ventilation (PPV) in order to increase the efficiency of the firefighting process and to reduce the adverse effects of fires. The tests have been carried out in unused buildings, providing real conditions. Practical application has been examined in order to reduce the harmful effects of closed-space fires and to provide guidance for professional use. This research based on observations and experiments contributes to enhancing fire safety.

<p&amp... more <p>Between the Ion Mincu University of Architecture and Urbanism in Bucharest, Romania and the Szechenyi Istvan University in Gyor, Hungary a cooperation agreement was concluded between the first and fourth author regarding disaster management. A first step was taken in January 2020 starting the reciprocical visits by a visit of the third author to the Romanian university. Exchange encompassed participation to master level courses at the Master Urban Design (urban prospective: urban vulnerability and protection of localities against risks, the later taught by the second author, who is also a titular member of the doctoral school) and a lecture at the doctoral school with discussions moderated by the first and third authors. The conclusions were discussed with the master students as well. The innovative in the cooperation is that it regards how urban planners can contribute to disaster management and infrastructures in a field where they can best plan. Master students learn how to design urban projects while doctoral candidates do research in this, and are thereof complementary. Cooperation will continue by various national and bilateral schemes. This contribution shows the conclusions of the discussions.</p>

International Journal of GEOMATE, 2018

The City of Jeddah in Saudi Arabia is expanding rapidly, in terms of new buildings and increasing... more The City of Jeddah in Saudi Arabia is expanding rapidly, in terms of new buildings and increasing population. The rapid urbanization leads to higher risk from seismic events; even in areas of moderate seismicity such as this city. The present study addresses the rapid evaluation of a large number of buildings in Jeddah involving steps to determine hazard, assessing building stock, and computing vulnerability with a scoring method from FEMA 155. Two districts were selected for investigation based on a cluster analysis applied to population and building data from the local municipality. One selected district was a contemporary developed urbanized area, and the other was a more traditional area. Such selection offered the possibility to compare vulnerability of buildings built according to different seismic codes and to make assumptions about the rest of the city based on typical structures of districts. The basic structural score was determined considering the building structure and moderate seismicity of the region using score modifiers, e.g. vertical irregularity score modifier; soil score modifier assuming sabkahs. The results of the investigation reveal a different level of vulnerability and areas where intervention is needed. The method can be applied for further analysis of the city.

Acta Technica Jaurinensis, 2015

This paper presents the process of earthquake risk analysis from the probabilistic determination ... more This paper presents the process of earthquake risk analysis from the probabilistic determination of seismic hazard and local site effects, through the evaluation of building vulnerability to an event resulting in seismic risk maps. These results can then serve as useful tools for decision makers and insurance companies, and can be applied directly to overall risk management plans of cities. Finally a case study for seismic risk assessment performed in city of Győr is presented, taking into account local site effects based on response analysis with more than 6000 realizations and rapid visual screening of 5000 building to obtain the seismic risk.

<p>One of the critical research areas of engineering science is the assessment of e... more <p>One of the critical research areas of engineering science is the assessment of existing buildings' seismic vulnerability. Since the existing building stock is made up of structures that were built before design codes were developed, taking into account low or moderate design codes may make them vulnerable to an upcoming earthquake. To this end, rapid visual screening (RVS) methods can be used to identify building vulnerability before or after an earthquake. A number of RVS methods have been developed; however, it's critical to assess their accuracy in terms of their ability to reliably classify the state of building vulnerability. Therefore, this study offers an application of conventional RVS methods (FEMA P-154, RISK-UE Project, JBDPA), as well as a comparison of the results of the building safety level classification of 20 existing reinforced concrete buildings from Győr, Hungary. Pushover analysis was used as a detailed vulnerability assessment technique in addition to the RVS methods to evaluate one more reinforced concrete building. The findings of pushover analysis and RVS methods are contrasted to demonstrate how effectively RVS methods can be utilized to determine building vulnerability. Additionally, the findings of this study can be used to select an RVS method for commencing a pre-earthquake building assessment of the existing reinforced concrete building stock in the investigation area.</p> <p><strong>Keywords:</strong> earthquake; existing buildings; vulnerability assessment; rapid visual screening; machine learning; building damage state</p>

Applied sciences, Oct 23, 2023

DOAJ (DOAJ: Directory of Open Access Journals), Sep 1, 2022

Our team has compiled Seismotectonic hazard map of Hungary. One of the main contents of the map a... more Our team has compiled Seismotectonic hazard map of Hungary. One of the main contents of the map are Eurocode 8 categories whose production steps are described here in detail. In engineering, site response to earthquakes has been classified to national and international standards. Eurocode 8 standard is partly based on Vs30 that is the time averaged shear-wave velocity in the uppermost 30 m sediment. We have compiled 67 Vs30 measurements and collected 103 Vs30 values from PhD theses and industrial reports. The values could be divided to soil class A–D of Eurocode 8 which are defined by Vs30 thresholds. The special soil class E (hard rock beneath 5–20 m thick loose sediment) needed a deeper investigation. The Vs trend was plotted and plots with obvious knickpoint has been analysed further. In case of one knickpoint in Vs trend two-layered model was used. We were defined the thickness and the theoretical Vs30 of the upper and the lower strata. In case if the site fit to class E, original Eurocode 8 class have been overwritten. Other advantage of the extrapolation of Vs trend of the uppermost strata is to derive the theoretical Vs30 of the given geomorphological feature if its sediment would fill up the whole 30 m.In Hungary only the youngest and lowest level of alluvial and lacustrine features fall into the most critical class D. Therefore that features have been mapped. In case of the youngest sediment’s thickness was not exceeded 20 m in each places, that site would classified as „shallow D” which is not a Eurocode 8 soil class. This process could be done using the borehole database of Geomega Ltd. Classification of soil class E have derived using the same method: thousands of borehole data have been checked to delineate the margin of the categories around the rock outcrops. For soil classes A–D topographical slope – Vs30 relation has established. For Hungary, we recommend to use 0.3%, 3% and 11% as topographical slope barriers between soil classes D-C-B-A (in advance).Secondly, active faults were mapped using the methodology described by the European Facilities for Earthquake Hazard and Risk. Third, earthquake database was use to present area affected by frequent ground motions. We have divided the database to historical and to instrumental detections due to their differences in the accuracy and reliability of magnitude and epicentre location.Historically Komárom-Oroszlány-Balatonfő line was most affected by earthquakes. Our map revealed that in the Middle Hungarian Shear Zone consists of still active fault lines. Some spots are affected by densely located small earthquakes such as the neighbourhood of Zalaszengrót, Répcelak, Nagyigmánd, the DIósjenő fault, Heves, Csepel, Jászberény, Nagykanizsa, Nagyatád, Pincehely, Szabadszállás, Kecskemét, and Miskolc. In almost all cases the most critical soil class D can be found in the neighbourhood of mentioned sites, while class E appears only in some locations.The research project was supported by the National Research, Development and Innovation Office of Hungary (2018-1.2.1-NKP-2018-00007). Map can be downloaded among others and vector data can be requested at Geomega website (www.geomega.hu).

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

Acta Geodaetica et Geophysica, 2015

Determination of shear-wave (S-wave) velocity profiles is the first step in seismic hazard assess... more Determination of shear-wave (S-wave) velocity profiles is the first step in seismic hazard assessment of a town, because the dynamic parameters of local soil types are vital for seismic response analysis of a specific area in order to determine the local soil effect in a case of a seismic event for seismic risk analysis. S-wave velocity profiles have been determined for many areas within Gy} or. Extensive use of historical boring logs allowed for correlations and reasonable extrapolation of soil performance throughout the area. This has led to a pattern of soil layer distributions and delineates several different soil zones for Gy} or. Keywords Local site effect Á Shear-wave velocity profile Á Dynamic properties of soils 1 Introduction Research in earthquake hazard mitigation has focused on evaluating possible damage scenarios for different magnitude events (Luco et al. 2007; Committee on National Earthquake Resilience 2011). Although seismic events are rare in many places, they are characterized by high exposure and their economic and social effects cannot be neglected.

Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous r... more Snow avalanches cause danger to human lives and property worldwide in high-altitude mountainous regions. Mathematical models based on past data records can predict the danger level. In this paper, we are proposing a neural network model for predicting avalanches. The model is trained with a quality-controlled sub-dataset of Swiss Alps. Training accuracy of 79.75% and validation accuracy of 76.54% have been achieved. Comparative analysis of neural network and random forest models concerning metrics like precision, recall, and F1 has also been carried out. 1. Introduction Accurate prediction of snow avalanches can help ensure people's safety in snow-covered regions. Many countries still depend on human experts to analyse meteorological data to forecast avalanche warnings. The major hurdle in developing machine learning models is the lack of sufficient and reliable data. This issue has been resolved to a great extent by the WSL Institute of Snow and Avalanche Research, Switzerland, by collecting 20 years of data in avalanche forecasting. This data set has been further refined with quality control by experts. The dataset combines different feature sets with meteorological variables. This unique dataset has enabled experimentation with machine learning models like neural networks and compared its performance with the random forest machine learning technique. This paper is organized as follows. Related literature is briefly overviewed in Section II. The dataset used for the training of neural networks is described in Section III. After that, in Section IV, we explain the neural network model, tuning of hyperparameters, and evaluation metrics. Random Forest machine learning method details applied to the same dataset are described in Section V. Results from both methods are compared and analysed in Section VI. The paper is concluded in Section VII. 2. Related Work Many countries face snow avalanche hazards with snow-clad mountains. It affects people, facilities, and properties. The impact of snow avalanches on living, work, and recreation in Canada is well documented (Sethem et. al., 2003). Every country

<p>One of the critical research areas of engineering science is the assessment of e... more <p>One of the critical research areas of engineering science is the assessment of existing buildings' seismic vulnerability. Since the existing building stock is made up of structures that were built before design codes were developed, taking into account low or moderate design codes may make them vulnerable to an upcoming earthquake. To this end, rapid visual screening (RVS) methods can be used to identify building vulnerability before or after an earthquake. A number of RVS methods have been developed; however, it's critical to assess their accuracy in terms of their ability to reliably classify the state of building vulnerability. Therefore, this study offers an application of conventional RVS methods (FEMA P-154, RISK-UE Project, JBDPA), as well as a comparison of the results of the building safety level classification of 20 existing reinforced concrete buildings from Győr, Hungary. Pushover analysis was used as a detailed vulnerability assessment technique in addition to the RVS methods to evaluate one more reinforced concrete building. The findings of pushover analysis and RVS methods are contrasted to demonstrate how effectively RVS methods can be utilized to determine building vulnerability. Additionally, the findings of this study can be used to select an RVS method for commencing a pre-earthquake building assessment of the existing reinforced concrete building stock in the investigation area.</p> <p><strong>Keywords:</strong> earthquake; existing buildings; vulnerability assessment; rapid visual screening; machine learning; building damage state</p>

Sustainability, Dec 6, 2022

This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY

Geosciences

In order to prevent possible loss of life and property, existing building stocks need to be asses... more In order to prevent possible loss of life and property, existing building stocks need to be assessed before an impending earthquake. Beyond the examination of large building stocks, rapid evaluation methods are required because the evaluation of even one building utilizing detailed vulnerability assessment methods is computationally expensive. Rapid visual screening (RVS) methods are used to screen and classify existing buildings in large building stocks in earthquake-prone zones prior to or after a catastrophic earthquake. Buildings are assessed using RVS procedures that take into consideration the distinctive features (such as irregularity, construction year, construction quality, and soil type) of each building, which each need to be considered separately. Substantially, studies have been presented to enhance conventional RVS methods in terms of truly identifying building safety levels by using computer algorithms (such as machine learning, fuzzy logic, and neural networks). This...

Earthquake Resistant Engineering Structures X, 2015

Assessment and management of earthquake risk requires several disciplines and different aspects t... more Assessment and management of earthquake risk requires several disciplines and different aspects to evaluate. Based on mathematical calculations, risk is the product of hazard and vulnerability. Considering earthquakes, the process of risk evaluation consists of several steps performed in parallel. One line of the research deals with the seismicity of the area by identifying potential sources, usually along fault systems, for probabilistic hazard assessment. The next step focuses on the simulation of strong ground motions for regional hazard assessment. Ground motion attenuation models are calibrated based on observed strong motion recordings and on theoretical computations taking into account realistic models for wave propagation as well as possible seismic source zones. Further research takes into account local site effects and will result in a microzonation for the evaluated area based on local soil conditions and field testing of the soil properties. Another line of research focuses on the built environment. This paper presents the case study for the city of Győr, focusing on the seismic hazard assessment of the typical soil types along the Danube. Extensive use of historical boring logs allowed for correlations and reasonable extrapolation of soil performance throughout the area. This has led to a pattern of soil layer distributions and delineates of several different soil zones for Győr. Compared to more simplified profile standards from building codes, it can be clearly seen that the different levels of hazard concerning Győr are apparent, even though almost all of the region belongs to soil category C based on EC8.

<p>Being part of the expert team sent by Hungarian National Directorate General for... more <p>Being part of the expert team sent by Hungarian National Directorate General for Disaster Management to Tirana after the 6.4 magnitude earthquake in 2019, and experienced the fear of residents living in slightly or heavily cracked buildings raised the question of preparedness to against seismic events in other moderate seismic regions such as Hungary. Recent earthquakes within the moderate range have proved that moderate seismicity does not necessarily equate to moderate damage suffered.</p><p>Vulnerability to earthquakes has increased due to extending urban areas. This paper presents lessons learnt after medium-sized earthquakes and examines the adaptability of measures taken afterwards and the possibility to apply these methods to other regions in Hungary and throughout Europe where the seismic hazard is not great, but cannot be ignored. To reduce the potential damage, a comprehensive assessment of the seismic risk followed by a package of relevant remedial measures is needed. Methods developed for Hungary is presented compared to methods applied in other regions to determine local site effects, vulnerability, and preparedness, being the main components responsible to risks. Based on the results, engineers can better plan to make improvements to infrastructure, and authorities can better plan for emergency activities in case of a seismic event.</p>

International Journal of GEOMATE, 2018

Closed-space fires often occur in Hungary, so it is necessary to examine the effects of fires on ... more Closed-space fires often occur in Hungary, so it is necessary to examine the effects of fires on building structures, taking into account Hungarian architectural features. Fires inside the buildings are characterized by intense heat development and smoke generation that can cause permanent damage to the building structures. Heat and smoke extraction during fire extinguishing is based usually on natural ventilation. Not only being a non-effective process also makes it more difficult to accomplish firefighting tasks. Experiments in this research have been conducted with mobile positive pressure ventilation (PPV) in order to increase the efficiency of the firefighting process and to reduce the adverse effects of fires. The tests have been carried out in unused buildings, providing real conditions. Practical application has been examined in order to reduce the harmful effects of closed-space fires and to provide guidance for professional use. This research based on observations and experiments contributes to enhancing fire safety.

<p&amp... more <p>Between the Ion Mincu University of Architecture and Urbanism in Bucharest, Romania and the Szechenyi Istvan University in Gyor, Hungary a cooperation agreement was concluded between the first and fourth author regarding disaster management. A first step was taken in January 2020 starting the reciprocical visits by a visit of the third author to the Romanian university. Exchange encompassed participation to master level courses at the Master Urban Design (urban prospective: urban vulnerability and protection of localities against risks, the later taught by the second author, who is also a titular member of the doctoral school) and a lecture at the doctoral school with discussions moderated by the first and third authors. The conclusions were discussed with the master students as well. The innovative in the cooperation is that it regards how urban planners can contribute to disaster management and infrastructures in a field where they can best plan. Master students learn how to design urban projects while doctoral candidates do research in this, and are thereof complementary. Cooperation will continue by various national and bilateral schemes. This contribution shows the conclusions of the discussions.</p>

International Journal of GEOMATE, 2018

The City of Jeddah in Saudi Arabia is expanding rapidly, in terms of new buildings and increasing... more The City of Jeddah in Saudi Arabia is expanding rapidly, in terms of new buildings and increasing population. The rapid urbanization leads to higher risk from seismic events; even in areas of moderate seismicity such as this city. The present study addresses the rapid evaluation of a large number of buildings in Jeddah involving steps to determine hazard, assessing building stock, and computing vulnerability with a scoring method from FEMA 155. Two districts were selected for investigation based on a cluster analysis applied to population and building data from the local municipality. One selected district was a contemporary developed urbanized area, and the other was a more traditional area. Such selection offered the possibility to compare vulnerability of buildings built according to different seismic codes and to make assumptions about the rest of the city based on typical structures of districts. The basic structural score was determined considering the building structure and moderate seismicity of the region using score modifiers, e.g. vertical irregularity score modifier; soil score modifier assuming sabkahs. The results of the investigation reveal a different level of vulnerability and areas where intervention is needed. The method can be applied for further analysis of the city.

Acta Technica Jaurinensis, 2015

This paper presents the process of earthquake risk analysis from the probabilistic determination ... more This paper presents the process of earthquake risk analysis from the probabilistic determination of seismic hazard and local site effects, through the evaluation of building vulnerability to an event resulting in seismic risk maps. These results can then serve as useful tools for decision makers and insurance companies, and can be applied directly to overall risk management plans of cities. Finally a case study for seismic risk assessment performed in city of Győr is presented, taking into account local site effects based on response analysis with more than 6000 realizations and rapid visual screening of 5000 building to obtain the seismic risk.

Sustainability, 2022

Developments in the field of earthquake engineering over the past few decades have contributed to... more Developments in the field of earthquake engineering over the past few decades have contributed to the development of new methods for evaluating the risk levels in buildings. These research methods are rapid visual screening (RVS), seismic risk indexes, and vulnerability assessments, which have been developed to assess the levels of damage in a building or its structural components. RVS methods have been proposed for the rapid pre- and/or post-earthquake screening of existing large building stock in earthquake-prone areas on the basis of sidewalk surveys. The site seismicity, the soil type, the building type, and the corresponding building characteristic features are to be separately examined, and the vulnerability level of each building can be identified by employing the RVS methods. This study describes, evaluates, and compares the findings of previous investigations that utilized conventional RVS methods within a framework. It also suggests the methods to be used for specific goal...