Aaltodoc

Aaltodoc is the institutional repository of Aalto University.

Aaltodoc has a new updated appearance. Instructions for searching and more information is available here.

Otakaari 1 grandhall. Photo: Esa Kapila
 

Communities in Aaltodoc

Select a community to browse its collections.

Now showing 1 - 13 of 13

Recent Submissions

Item
Perturbaatioteoria
(2023-09-11) Eklund, Linnea; Hakula, Harri; Perustieteiden korkeakoulu; Hakula, Harri
Työssä esitellään perturbaatioteoria: sen keskeinen tarkoitus, tieteellinen motivaatio työn tekemiseen sekä aiheeseen keskeisesti liittyvät Gerschgorinin teoreema että ominaisarvojen ja-vektorien derivaatat. Työ keskittyy erityisesti matriisiperturbaatiooon, jonka tavoitteena on vastata kysymykseen, miten matriisin ominaisarvot, niitä vastaavat ominaisvektorit sekä muut matriisiin liittyvät ominaisuudet muuttuvat, kun alkuperäinen matriisi koh taa pientä perturbaatiota eli häiriötä. Perturbaatioanalyysillä saadaan selville joko perturbaatiolaajennus tai perturbaatioraja. Laajennus arvioi funktion perturbaation määrää, kun tiedetään perturbaatio sen muuttujassa. Perturbaatiorajat rajaavat perturbaatoin määrän yksittäisessä muuttujassa, rajoja voidaan luokitella erilaisiin ryhmiin. Eri matriisiluokat käyttäytyvät eri tavoin altistuessaan perturbaatiolle, minkä takia eri matriisiluokille on omat yksilölliset perturbaatiorajat ja-laajennukset. Teorian suurin rajoite on perturbaation koko. Teoria toimii parhaiten, kun per turbaatio tai muutokset ovat riittävän pieniä. Jos perturbaatio on liian suuri, ap proksimaatioiden tarkkuus heikkenee. Ominaisarvojen derivaatat ja Gerschgorinin teoreema eivät ole suoraan pertur baatiorajoja, mutta niiden avulla saadaan tietoa jota voidaan hyödyntää teorian käytössä. Ominaisarvojen derivaatat kuvaavat niiden muutosta matriisin paramet reihin tai muuttujiin, esimerkiksi perturbaatioparametriin ϵ. Derivaatoilla voidaan arvioida ominaisarvojen muutosta perturbaation vaikutuksesta. Gerschgorinin teo reeman avulla saadaan selville ominaisarvojen summitainen sijainti kompleksitasolla, ominaisarvot sijaitsevat Gerschgorinin kiekkojen yhdisteessä. Perturbaatioteoriaa hyödynnetään insinööritieteissä monilla eritieteen aloilla. Suurten ominaisarvo-ongelmien numeerinen ratkaisu on usein haastavaa ja teoria tar joaa hyvän numerisen menetelmän ongelmien ratkaisemiseksi, kun matriisiin tehdään pieniä muutoksia. Eniten omninaisarvo-ongelmia ratkaistaan rakennusdynamiikassa, jossa niitä käytetään rakennusten, siltojen ja muiden rakenteiden suunnittelussa, rakenteiden käyttäytymisen ja värähtelyjen analysoimiseksi.
Item
Modelling activity networks of an organization
(2024-05-24) Kanerva, Pyry; Hakula, Harri; Perustieteiden korkeakoulu; Hakula, Harri
Knowledge work is most often conducted using productivity and communication software tools. Typically, multiple people are involved in every independent activity, and most activities are linked to other activities conducted by the organization. Leading and managing work that mostly occurs within multiple software tools is difficult. Fingertip software creates a leadership layer to Microsoft Teams, enabling workers and leaders to maintain a grasp of the ongoing activities of the business. This work explores the activity networks of a small information technology company through the lens of network theory. The study of these networks reveals how the activity networks of an organization can often exhibit hub-and-spoke structures and sometimes small-world phenomena. The study can, however, be skewed by the connection between these distinct activities might exist in the real working environment but not in the Fingertip data. The users of tools and their limitations determine the extent to which data compiled by the tool represent the real world. This work was motivated by the multiyear conversation between D.Sc. (Tech.) Harri Hakula and Dr. Jere Koskinen about team efficacy.
Item
Growth and Characterisation of chromium selenide
(2024-05-22) Lemmetty, Heidi; Drost, Robert; Perustieteiden korkeakoulu; Liljeroth, Peter
Two dimensional (2D) materials show a wide range of interesting properties, including superconductivity, ferroelectricity, antiferromagnetism and ferromagnetism. Espe- cially transition metal dichalcogenides (TMDC), a type of 2D-materials, have been researched in the recent years. CrSe2 is a TMDC that has shown antiferromagnetic properties which could have applications in spintronics and magnetoelectronics. The magnetic properties of CrSe2 have been observed to be tunable by for example applying strain, using hydrogenation and adjusting layer thickness. Molecular beam epitaxy (MBE) is a growth method to produce clean samples of 2D-materials for research applications. MBE is a slow growth process that operates in ultra high vacuum (UHV) where growth materials are evaporated. Because of the conditions, clean and well structured samples can be produced. In the experiments done for this thesis, MBE was used to grow monolayers of chromium selenides. Samples were initially analysed using scanning electron microscopy (SEM) to determine the island shapes. Once desired islands and coverage were obtained, the grown phase was determined based on analysis done using x-ray photoelectron spectroscopy (XPS) and scanning tunneling microscope (STM). The best results were achieved using a seeding process in the growth. Two different types of islands were produced on highly ordered pyrolytic graphite (HOPG). On the sample clear triangular islands were observed along with rounder islands of lesser quality. The clearer islands were analysed with STM and the data was used to determine the lattice constant of the material. Based on the lattice constant and similar height profiles of the two materials, the islands are likely two different phases of CrSe2. As the sample was in contact with air before STM measurements, especially the rounder islands contained impurities which could have been avoided with selenium capping.
Item
Molecular descriptor engineering for machine learning predictions in atmospheric science
(2024-05-16) Lind, Linus; Sandström, Hilda; Perustieteiden korkeakoulu; Rinke, Patrick
Atmospheric organic compounds form complex mixtures, where new compounds may form via chemical reactions, increasing the number of unique molecular species and their complexity. These compounds may exist in gas phase and in particulate matter phase, suspended as aerosols. Organic aerosols are a particular focus of research in atmospheric science, since they affect radiative forcing, air quality, and other chemical and physical processes such as cloud formation in the atmosphere. Key physico-chemical properties of a molecule that determine aerosol formation are saturation vapour pressure and equilibrium partition coefficients, but unfortunately these are difficult to measure experimentally for atmospheric compounds. Thus, the use of machine learning methods for predicting these properties are on the rise, increasing the demand for new ways to store molecular data in a machine-readable format, in the form of a molecular descriptor. In this thesis, a new binary encoded molecular descriptor was developed for machine learning purposes: the BESPE-MACCS descriptor. It is based on the Molecular Access System (MACCS) descriptor and a molecular substructure enumeration method called binary encoded SMARTS pattern enumeration (BESPE). The aim of this thesis was to improve the applicability of the MACCS descriptor for predicting properties of atmospherically relevant organic compounds. The performance of the newly developed BESPE-MACCS descriptor was evaluated using a kernel ridge regression machine learning model, which was used in a previous study that compared prediction accuracies obtained by the model with different descriptors. In prediction accuracy, measured by mean absolute error, the BESPE-MACCS descriptor had similar or better performance compared to descriptors used in previous research within statistical significance, while having a much smaller file size and computational expense. Additionally, the BESPE-MACCS descriptor is human-interpretable, customisable, and simple, thus easy to develop further, which makes it a promising tool for studying organic aerosols in atmospheric science.
Item
Machine Learning in Magnetospheric Physics Time Series Data
(2024-04-26) Clegg, Daniel; Weigt, Dale; Perustieteiden korkeakoulu; Korpi-Lagg, Maarit
Magnetospheric physics today relies heavily on time series data describing the complex dynamics between different magnetic regions and the particle populations within them. Attributed to the availability of large volumes of time series data, the process of manually labelling boundary crossings by hand has shifted to machine learning induced automation. However, few implementations of intricate machine learning based boundary crossing detection methods exist. Two recent studies have applied threshold, deep learning and unsupervised learning models trained on time series data collected by the Cassini probe as it was exploring the magnetosphere of Saturn. In this study a comprehensive comparison between the undertaken approaches is conducted, resulting in an evaluation of 8 total models. Taking into consideration the application context of each model identified by the authors, the best performing model is based on Matrix Profile, scoring a recall value of 63.6% at a precision above 80%. High runtimes, low F1-scores and overfitting models limit current machine learning implementations from being applied locally on-board future space missions. Henceforth, future machine learning model improvements and/or implementations for boundary crossing detection should aim to fully leverage the extensive data available, constructing training sets spanning all years while maintaining computational efficiency.