12345 101 - 150 of 211
rss atomLink to result list
Permanent link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
  • Ismailoglu, Nilab
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Antenna Selection and Bit Allocation Strategies for Generalized Spatial Modulation2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Spatial Modulation (SM) is a novel transmission technique for Multiple InputMultiple Output (MIMO) systems. In SM, a block of information bits is dividedinto two, and mapped to constellation symbols and spatial symbols. Spatialinformation is carried by the active transmit antenna index for each transmittedblock. Generalized Spatial Modulation (GSM) overcomes the single activetransmit antenna constraint of SM, and lets multiples of transmit antennas tobe activated at the same time. This allows GSM to increase the amount ofspatial symbols and the spectral efficiency with respect to SM despite of havingthe same number of transmit antennas. However, since the receiver needs thechannel information to detect the transmitted symbols, performance of GSM isaffected significantly under unfavorable channel conditions. In this thesis, antennaselection and bit allocation strategies are presented to improve the errorperformance of GSM systems with the Channel State Information (CSI) knownat the receiver.Since there can be more possible antenna combinations than the amountthat can be utilized, employing an antenna selection strategy is necessary inGSM systems. In order to improve average Bit Error Rate (BER) of GSMsystems, a novel antenna selection method is proposed. The proposed methodmaximizes the Euclidean distance of the selected antenna combinations with theChannel State Information at the Receiver (CSIR). By making use of statisticalproperties of the channel, a heuristic which decreases the search domain for theantenna selection method is presented.Another factor translating into bit error performance of communication systemsis bit allocation. The proposed bit allocation strategies aim at minimizingthe Hamming distance for the symbols that are neighboring in Euclidean spacewith the given CSI. The first approach for bit allocation utilizes set partitioningof the signal constellation, the second is based on the Delaunay tessellation ofthe signal constellation.The strategies presented in this thesis for both antenna selection and bit allocationare analyzed in terms of performance and in computational complexity.The algorithms for the implementation of each proposed strategy are describedin detail. The BER performance of the proposed algorithms is demonstrated byMonte Carlo simulations. They are compared in error performance and computationalcomplexity to each other and to the conventional approaches in theliterature.

  • Panagiota Mastoropoulou, Emmeleia
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Enhancing Deep Active Learning Using Selective Self-Training For Image Classification2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    A high quality and large scale training data-set is an important guarantee to teach an ideal classifier for image classification. Manually constructing a training data- set  with  appropriate  labels  is  an  expensive  and  time  consuming  task.    Active learning techniques have been used to improved the existing models by reducing the  number  of  required  annotations.    The  present  work  aims  to  investigate the  way  to  build  a  model  for  identifying  and  utilizing  potential  informative and  representativeness  unlabeled  samples.    To  this  end,  two  approaches  for deep image classification using active learning are proposed, implemented and evaluated.  The two versions of active leaning for deep image classification differ in  the  input  space  exploration  so  as  to  investigate  how  classifier  performance varies  when  automatic  labelization  on  the  high  confidence  unlabeled  samples is  performed.    Active  learning  heuristics  based  on  uncertainty  measurements on low confidence predicted samples,  a pseudo-labelization technique to boost active  learning  by  reducing  the  number  of  human  interactions  and  knowledge transferring  form  pre-trained  models,  are  proposed  and  combined  into  our methodology.  The experimental results on two benchmark image classification data-sets  verify  the  effectiveness  of  the  proposed  methodology.    In  addition, a  new  pool-based  active  learning  query  strategy  is  proposed.     Dealing  with retraining-based algorithms we define a ”forgetting event” to have occurred when an  individual  training  example  transitions  the  maximum  predicted  probability class over the course of retraining. We integrated the new approach with the semi- supervised learning method in order to tackle the above challenges and observedgood performance against existing methods.

  • Torre, Ilaria
    et al.
    University of Plymouth.
    White, Laurence
    University of Plymouth.
    Goslin, Jeremy
    University of Plymouth.
    Behavioural mediation of prosodic cues to implicit judgements of trustworthiness2016In: Proceedings of the eighth International Conference on Speech Prosody 2016, ISCA , 2016Conference paper (Refereed)
  • Torre, Ilaria
    et al.
    Trinity College Dublin.
    Carrigan, Emma
    Trinity College Dublin.
    McDonnell, Rachel
    Trinity College Dublin.
    Domijan, Katarina
    Maynooth University.
    McCabe, Killian
    ADAPT Research Centre.
    Harte, Naomi
    Trinity College Dublin.
    The Effect of Multimodal Emotional Expression and Agent Appearance on Trust in Human-Agent Interaction2019In: Motion, Interaction and Games, Association for Computing Machinery , 2019Conference paper (Refereed)
  • McGinn, Conor
    et al.
    Trinity College Dublin.
    Torre, Ilaria
    Trinity College Dublin.
    Can you Tell the Robot by the Voice?: An Exploratory Study on the Role of Voice in the Perception of Robots2019In: 14th ACM/IEEE International Conference on Human-Robot Interaction, HRI 2019, Daegu, South Korea, March 11-14, 2019, 2019, p. 211-221Conference paper (Refereed)
  • Torre, Ilaria
    et al.
    Trinity College Dublin.
    Carrigan, Emma
    Trinity College Dublin.
    McCabe, Killian
    ADAPT Research Centre.
    McDonnell, Rachel
    Trinity College Dublin.
    Harte, Naomi
    Trinity College Dublin.
    Survival at the Museum: A Cooperation Experiment with Emotionally Expressive Virtual Characters2018In: Proceedings of the 2018 on International Conference on Multimodal Interaction, 2018, p. 423-427Conference paper (Refereed)
  • Torre, Ilaria
    et al.
    Trinity College Dublin.
    Goslin, Jeremy
    University of Plymouth.
    White, Laurence
    University of Plymouth.
    Zanatto, Debora
    University of Plymouth.
    Trust in artificial voices: A “congruency effect” of first impressions and behavioural experience2018In: Proceedings of APAScience ’18, 2018Conference paper (Refereed)
  • El Haddad, Kevin
    et al.
    University of Mons.
    Torre, Ilaria
    University of Plymouth.
    Gilmartin, Emer
    Trinity College Dublin.
    Çakmak, Hüseyin
    University of Mons.
    Dupont, Stéphane
    University of Mons.
    Dutoit, Thierry
    University of Mons.
    Campbell, Nick
    Trinity College Dublin.
    Introducing AmuS: The Amused Speech Database2017Conference paper (Refereed)
    Abstract [en]

    In this paper we present the AmuS database of about three hours worth of data related to amused speech recorded from two males and one female subjects and contains data in two languages French and English. We review previous work on smiled speech and speech-laughs. We describe acoustic analysis on part of our database, and a perception test comparing speech-laughs with smiled and neutral speech. We show the efficiency of the data in AmuS for synthesis of amused speech by training HMM-based models for neutral and smiled speech for each voice and comparing them using an on-line CMOS test.

  • Ye, Jingzhi
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Two-stage Soldering Defect Detection with Deep Learning2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    In  electronics  industry,  quality  control  of  soldering  points  on  printed  circuitboards (PCB) is an important topic.  Soldering points are usually inspected by the operator or by automatic optical inspection (AOI) techniques.This thesis proposes an AOI method for soldering fault detection based on deep learning.  There are two stages of the method.  The first stage is to detect all the soldering points from the PCB image, and the second stage is to classify them by their types of defect.  According to our results on the most common defect type, the method is a feasible solution to detect soldering faults and ithas great potential to be deployed to the actual production environment.

  • Hu, Ao
    KTH, School of Electrical Engineering and Computer Science (EECS).
    On the Influence of Scattering From a Power-pole on an Airport Radar2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Power-poles are one of the common massive conducting structures in cities and countryside. To fulfill the remarkable increased electric power demand nowadays, the number and the size of power-poles are also growing over the past few decades.  The influence of power-pole on electromagnetic  wave  propagation  is  a  potential  noise  source  for  nearby  radio  devices  like radars.   This master thesis project is aiming to analysis the interference of a case where an Amplitude Modulated (AM) signal emitted. We consider here the Stockholm Arlanda Airport’s Very High Frequency (VHF) Omnidirectional Range (VOR) radar scattering by a power-pole 2.6km away.A Method of Moment technique is used to solve the scattering problem.  We modeled the power-pole on a blueprint from Svenska Kraftnät. It is meshed in small triangles by AutoCAD and  GMSH  software,  and  on  that  mesh  the  Rao-Wilton-Glisson  (RWG)  basis  function  are formed.  A MoM code developed by Makarov determines the bi-static scattering pattern of the power-pole.  Four main models have been considered, a one-side power-pole model and the same model with the ground plane, as well as two sides power-pole model and the model with the ground plane.  We have assumed that the incident field on the power-pole is a plane wave and that the ground is an infinite PEC surface.  The result is presented by the Interference to Signal ratio (ISR) of an airplane receiver when it is flying toward the airport for landing.By the end of the project, we have shown for the considered model that a    50    15dB level of interference is estimated for the interference to signal ratio.  This result then suggests that a power-pole may have a little signal interference towards the VOR system radar wave emit awayfrom 2.6km.

  • Papadopoulos, Georgios
    KTH, School of Electrical Engineering and Computer Science (EECS).
    State-of-the-art DC Current Interruption Concept using SiCMOSFETs2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    The transportation is in transition to the electrification era where the main source of energy for thevehicles is supplied by high energy propulsion batteries with voltages above 400 V. Currently the majority of the batteries are lithium-ion batteries and it is crucial to protect them from shortcircuits in order to ensure the safe operation in case of a vehicle crash or misuse of the vehicle. This Master Thesis deals with a DC-current interruption method using SiC MOSFETs which is meant to be used as protection device in the propulsion battery of heavy vehicles.Initially in this report, the contemporary design of the battery junction box is discussed along with  the existing overcurrent protection at the battery pack. The motivation of using SiC MOSFETs follows and a conducted pre-study about using SiC MOSFETs on the pre-charge branch of the battery  junction box is mentioned. Regarding the literature study, it is subdivided into two subsection, the DC current interruptions methods where the contemporary solutions for protecting against overcurrents are mentioned and the SiC MOSFETs where information about their structure, modelling, reliability and capabilities are presented. Afterwards, the importance of considering a multi-battery vehicle and increasing the number of parallel MOSFETs to reduce the power losses is pointed out. The modelling of the suggested protection device follows and the simulation circuits at LT Spice are presented. The features of self-triggering in case of a short-circuit and performing pre-charge of the vehicles DC-link capacitance are added to the protection device and the results from the simulation indicate its capability to disconnect the battery within nanoseconds ensuring the safe clearance of a shortcircuit.Experimental validation of the concept is provided by testing a prototype which have been designed and developed in the frame of this master thesis. Based on the findings of the thesis it can be concluded that the SiC solid state circuitbreaker constitutes a feasible solution which adds controllability in the system increasing significantly the speed of the protection system. However, it isnot currently applicable due to the high cost of SiC technology.

  • Xiong, Jiarui
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Active Learning for Semantic Segmentation2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Neural Networks have proven their capabililties in computer vision tasks. How- ever, their ability depends on the quantity and quality of data. Larger training dataset usually improves the performance of the trained model. This, however, comes at a cost. Collecting new data points for training and especially labelling and annotating them is a labour-intensive task which very often needs signifi- cant time and money. Therefore it is important to devote our time and money for collecting the data which makes a difference.  If there are already a lot of similar data points in our dataset, there is no point in collecting and annotat- ing one more similar to them.  Instead one might prefer to collect data which adds new “information” to the existing training dataset.In the segmentation task, labeling one image demands approximate one hour by the human. In this research project, the problem of informative data collection will be studied. The method will be used for collecting training data to train a deep learning- based semantic segmentation algorithm. To reduce the cost of preparing data, making our neural network learn the most informative data actively is an ur- gent need.  I investigate the method to measure the information contained in data and build the pipeline to improve the neural network performance in the segmentation task. Implementing details and current result will be presented.

  • Bilal Akhtar, Muhammad
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Optical Inspection for Soldering Fault Detection in a PCB Assembly using Convolutional Neural Networks2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Convolutional Neural Network (CNN) has been established as a powerful toolto automate various computer vision tasks without requiring any aprioriknowledge. Printed Circuit Board (PCB) manufacturers want to improve theirproduct quality by employing vision based automatic optical inspection (AOI)systems at PCB assembly manufacturing. An AOI system employs classiccomputer vision and image processing techniques to detect variousmanufacturing faults in a PCB assembly. Recently, CNN has been usedsuccessfully at various stages of automatic optical inspection. However, nonehas used 2D image of PCB assembly directly as input to a CNN. Currently, allavailable systems are specific to a PCB assembly and require a lot ofpreprocessing steps or a complex illumination system to improve theaccuracy. This master thesis attempts to design an effective soldering faultdetection system using CNN applied on image of a PCB assembly, withRaspberry Pi PCB assembly as the case in point.Soldering faults detection is considered as equivalent of object detectionprocess. YOLO (short for: “You Only Look Once”) is state-of-the-art fast objectdetection CNN. Although, it is designed for object detection in images frompublicly available datasets, we are using YOLO as a benchmark to define theperformance metrics for the proposed CNN. Besides accuracy, theeffectiveness of a trained CNN also depends on memory requirements andinference time. Accuracy of a CNN increases by adding a convolutional layer atthe expense of increased memory requirement and inference time. Theprediction layer of proposed CNN is inspired by the YOLO algorithm while thefeature extraction layer is customized to our application and is a combinationof classical CNN components with residual connection, inception module andbottleneck layer.Experimental results show that state-of-the-art object detection algorithmsare not efficient when used on a new and different dataset for object detection.Our proposed CNN detection algorithm predicts more accurately than YOLOalgorithm with an increase in average precision of 3.0%, is less complexrequiring 50% lesser number of parameters, and infers in half the time takenby YOLO. The experimental results also show that CNN can be an effectivemean of performing AOI (given there is plenty of dataset available for trainingthe CNN).

  • Ou, Shuyu
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Maximum Inverter AC Current Prediction Based on Junction Temperature Calculation2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Semiconductor devices are widely used in the automotive industry, and they are key components of the inverter and the converter in an electric vehicle. Thus, the concern of protecting the inverter and the converter from damaging operation has been raised. The junction temperature is one of the main considerations. It is directly related to the inverter power loss and overheat which can lead to fatigue or even failures. Therefore, to protect the semiconductor device from thermal runaway and apply active thermal control, the junction temperature must be obtained.To derive the junction temperature of an Insulated-Gate Bipolar Transistor (IGBT) and a diode, an instant model is proposed.   The instant model updates the junction temperature at the PWM switching frequency (around 2-5 kHz in this thesis). Compared with the traditional algorithm which calculates the average temperature over one fundamental cycle, the instant algorithm can show the instant temperature swing.  A high power IGBT module, FF1200R12IE5 Infineon, is selected to verify the algorithm with temperature results from IPOSIM and Power HIL tests. The temperature deviations for different cases are below 6 ◦C and the relative errors are below 10 %.With an accurate estimation of the junction temperature, the current limit is set to avoid that average temperature, maximum temperature and temperature swing exceed their limits. The currentlimits are derived from the curve/ surface fitting method.

  • Isakson, Alexander
    et al.
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Tunestam, Tobias
    KTH, School of Electrical Engineering and Computer Science (EECS).
    The Feasibility of Photovoltaics in Shadow Intensive Areas2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    The  shift  towards  renewable  energy  sources  is  ongoing,  and  with  scientific  advances  constantly being  made  within  the  area,  solar  panels  are  becoming  cheaper  and  more  efficient  by  the  day. Thus far, solar panels are almost exclusively being installed in areas where little to no shading is present since the shadows are a detriment to their performance.  The northern European climate also  presents  a  problem  in  tending  to  be  very  cloudy.   This  study  investigates  the  impacts  of shadows on the overall performance of a photo-voltaic system situated within an area surrounded by trees and containing a wind power.  This is done to investigate the feasibility of converting the wind  power  farm  to  a  wind/solar  hybrid  farm.   A  time  based  method  for  modeling  shadows  on a  two-dimensional  surface  is  developed  and  combined  with  a  method  for  investigating  irradiance data in order to find the irradiance of an area subject to shadowing.  A temperature and irradiance dependent model for calculating the power output of solar panels is studied and implemented in order  to  find  the  energy  output  potential  of  each  coordinate  in  the  two  dimensional  area  over  a year.  A basic economic analysis is performed to give grounds for justifying the implementation of solar  panels  in  the  area.  The  influence  of  solar  panels  shading  each  other  is  investigated  to  find the optimal spacing between them.  The results are then used to find at what coordinates the solar panels should optimally be placed.  Their power output is investigated and matched with electricity price data and an economic argument is made for installing the solar panels.  The area in question is subject to use by heavy duty vehicles for maintenance of the wind power turbines, therefore a few solutions for having the solar panels removable are discussed.

  • Gidofalvi, Gyözö
    et al.
    KTH, School of Architecture and the Built Environment (ABE), Urban Planning and Environment, Geoinformatics. KTH, School of Industrial Engineering and Management (ITM), Centres, Integrated Transport Research Lab, ITRL.
    Yang, Can
    KTH, School of Architecture and the Built Environment (ABE), Urban Planning and Environment, Geoinformatics.
    The potential of route based ERS network optimization2020In: TRA2020 Book of Abstracts: Proceedings of 8th Transport Research Arena TRA 2020, April 27-30, 2020, Helsinki, Finland / [ed] TRA, 2020, p. 1-12, article id 960Conference paper (Refereed)
    Abstract [en]

    The large scale deployment of Electric Road Systems (ERS) is a necessary and viable choice for reaching the emission reduction targets in the road-bound heavy freight sector. The per-kilometer infrastructure development cost of ERS is large, thus selecting segments that yield a high utility is important. According to a newly introduced concept, the electrification utility of a segment in a network is highly dependent on the freight routes-, the powertrain technology-, the energy supply and demand- and the transport loads of the vehicles as well as the topographic aspects- and traffic state of the road network. This paper explains these concepts and aspects and provides first empirical evidence about the potential of route based ERS network optimization that takes these aspects into consideration. Results show that the potential cost savings are up to 75%, which for national expressway networks is estimated to be in the range of 120M€ to 8,520M€.

  • Jiang, Frank
    KTH, School of Electrical Engineering and Computer Science (EECS).
    The Automated Vehicle Traffic Control Tower2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    This thesis presents and motivates the integration of automated vehicle traffic control towers (AVTCT) into transportation networks to support the operation of automated driving systems. Loosely analogous to air traffic control towers, the AVTCT provides a human-driven exception-handling layer to automated driving systems for safe operation in unexpected or uncertain, safety-critical scenarios.  In the thesis, a case-study of plausible scenarios is first presented that either pose difficulty for automated systems or require the supervision of human operators. An outline of the AVTCT concept and its different potential roles in the support of automated driving systems is then discussed.  Using reachability analysis, a concrete example from the case-study is formulated to show how the AVTCT would support an automated driving system that has reached an infeasibility in its automation. Finally, a demonstration of the example using an experimental implementation of the AVTCT and a scaledvehicle is presented.

  • Alatawi, Hatem
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Determining an Optimal Level of Power System Investments Under Large Scale Penetration of Solar Power in Saudi Arabia2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    The Saudi Arabia government set ambitious plans to adopt renewable energy sources in the energy mix as part of its strategy to diversify its economy away from oil. Ac- cording to the renewable energy project development office (REPDO), the total RES capacity installed would amount to 27.3 GW by 2023 and 58.7 GW by 2030.  Due to the geographic location of Saudi Arabia, solar energy is a promising renewable energy source. However, there are many challenges to achieving a future where so- lar generation represents a significant portion of the Saudi generation mix.  These concerns relate to the characteristics of solar PV (e.g., Variability, Aerosols, intermit- tency). As a result, measures should in place to take full advantage of the ambitiousplans.By modeling the generation of PV using real-time measurements, it is possible to quantify the potential energy produced (e.g., Power DC output from the PV panel). Also, through optimization techniques, it is possible to optimize future investments (e.g., capacity and transmission line investments) to minimize the costs while ensur- ing a reliable power system.  In this thesis, the model accounts for the variability in the hourly solar production for an entire year, by investing in the required capacity to meet the hourly demand and the necessary PV-operational reserves for multiple interconnected regions. Also, the model optimizes the investments over a given fu- ture growth in demand (electricity consumption) while accounting for the current generation mix, fuel prices, and PV deployment in each region.The model could be used to investigate multiple policies and their outcomes (e.g., fuel prices,  PV regional deployment,  PV capacity).   In this thesis,  two cases have been simulated based on the policymaker’s plans for adopting PV. The first case ex- amines 9.5 GW of installed PV capacity by 2023, while the second looks into 40 GW of installed PV capacity by 2030. The outcome is quantified by measuring the system costs and CO2 emissions.

  • Sanyang, Pa
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems.
    Web-based prototype for protecting controllers from existing cyber-attacks in an industrial control system2020Independent thesis Basic level (professional degree), 10 credits / 15 HE creditsStudent thesis
    Abstract [en]

    Industrial control system or ICS is a critical part of the infrastructure in society. An example of ICS is the rail networks or energy plants like the nuclear plant. SCADA is an ICS system following a hierarchical structure. Due to the fact that a control system can be very large, monitoring remote through networks is an effective way to do so. But because of digitalization ICS or SCADA systems are vulnerable to cyber attacks that can hijack or intercept network traffic or deny legitimate user services. SCADA protocols (e.g. Modbus, DNP3) that are prone to get attacks due to not being a secure protocol make a SCADA system even more vulnerable to attacks. The paper focuses on how to best protect the network traffic between an HMI as the client and a different controller as the server from attacks. The proposed solution, the prototype, is based on the reverse proxy server setup to protect controllers from the external network traffic. Only the reverse proxy server, or gateway server, can forward a client request to the intended controller. The gateway server, a web-based solution, will be the additional security layer that encrypts the payload in the application layer using TLS version 1.2 by using HTTPS protocol, thereby protect from usual security threats. The prototype went through a penetration testing of MITM (Based on ARP-poisoning), SYN flooding, slow HTTP POST attacks. And the result indicated that the prototype was vulnerable to SYN flooding and the network traffic was intercepted by the MITM. But from the Confidentiality-Integrity-Availability (C.I.A) criteria, the prototype did uphold the integrity and confidentiality due to the TLS security and successful mitigation of certain attacks. The results and suggestions on how to improve the gateway server security were discussed, including that the testing was not comprehensive but that the result is still valuable. In conclusion, more testing in the future would most likely showcase different results, but that will only mean to better the security of the gateway server, the network that the client and gateway server runs in and the physical security of the location where the client and gateway server is located.

  • Hermansson, Cecilia
    et al.
    KTH, School of Architecture and the Built Environment (ABE), Real Estate and Construction Management, Real Estate Economics and Finance.
    Sara, Jonsson
    Stockholm University, Stockholm Business School.
    Managing one's personal finances: Are women more overconfident than men?Manuscript (preprint) (Other academic)
    Abstract [en]

    We investigate gender differences in overconfidence in managing one’s personal finances and find that women are more overconfident than men in this context. Overconfidence decreases with educational level for both genders, however, the decrease is larger for men.

  • Azami Ghadim, Sohrab
    KTH, School of Technology and Health (STH), Medical Engineering, Medical Imaging.
    Utilizing Multi-Core for Optimized Data Exchange Via VoIP2016Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    In contemporary IT industry, Multi-tasking solutions are highly regarded as optimal solutions, because hardware is equipped with multi-core CPUs.  With Multi-Core technology, CPUs run with lower frequencies while giving same or better performance as a whole system of processing. This thesis work takes advantage of multi-threading architecture in order to run different tasks under different cores such as SIP signaling and messaging to establish one or more SIP calls, capture voice, medical data, and packetize them to be streamed over internet to other SIP agents. VoIP is designed to stream voice over IP. There is inter-protocol communication and cooperation such as between the SIP, SDP, RTP, and RTCP protocols in order to establish a SIP connection and- afterwards- stream media over the internet. We use the Microsoft COM technology in order to better the C++ component design. It allows us to design and develop code once and run it anywhere on different platforms. Using VC++ helps us reduce software design time and development time. Moreover, we follow software design standards setup by software engineers’ society. VoIP technology uses protocols such as the SIP signaling protocol to locate the user agents that communicate with each other. Pjsip is a library that allows developers to extend their design with SIP capability. We use the PJSIP library in order to sign up our own developed VoIP module to a SIP server over the Internet and locate other user agents. We implement and use the already-designed iRTP protocol instead of the RTP to stream media over the Internet. Thus, we can improve RTP packet delays and improve Quality of Service (QoS). Since medical data is critical and must not be lost, the iRTP guarantees no loss of medical data. If we want to stream voice only, we would not need iRTP, because RTP is a good protocol for voice applications. Due to the increasing Internet traffic, we need to use a reliable protocol that can detect packet loss of medical data. iRTP resolves the issue and leverages QoS. This thesis work focuses on streaming medical data and medical voice-calls using VoIP, even over small bandwidths and in high traffic periods. The main contribution of this thesis is in the parallel design of iRTP and the implementation of this very design in order to be used with Multi-Core technology. We do so via multi-threading technology to speed up the streaming of medical data and medical voice-calls. According to our tests, measurements, and result analyses, the parallel design of iRTP and the multithreaded implementation on VC++ leverage performance to a level where the average decrease in delay is 71.1% when using iRTP for audio and medical data instead of the nowadays applied case of using an RTP stream for audio and multiple TCPs streams for medical data .

  • Ashok, Joshua David
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems.
    The use of and attitudes to voice amplification in higher education institutions2020Independent thesis Advanced level (degree of Master (One Year)), 10 credits / 15 HE creditsStudent thesis
    Abstract [en]

    Background: Voice is still the most important tool used by teachers, but voice is varied and subject to problems. Teachers who lecture for long hours or have taught for most of their professional lives, face issues such as voice discomfort, chronic voice disorders, occupational voice disorder, stress, physiological and psychological problems. Due to issues such as poor acoustics in the classroom and absence of any installed voice amplification system, teachers might be subject to increase their vocal load to reach all the students in the classroom.

    Purpose: The aim of this study is to study the fundamental frequency (fo) and the strength Leq[dB(C)] of the teacher’s voice, with and without voice amplification system. 

    Method: Students from the preparatory year of their studies at KTH (CBH), were asked if they would like to participate in this study that researches on the voice health of teacher. Two days of teaching sessions were recorded using a microphone attached to the teacher’s mouth and three other static microphones placed in different parts of the room. The voice amplification system used in the study, was the one that was already in use at some of the classrooms at KTH. The students and the participating teacher were given questionnaires, towards the end of the teaching session to voice their opinion on the teaching voice, with and without the voice amplification system. 

    Results: The results from the data collected from the recordings, have shown that the SPL increased between 0.6 dB(C) and 5 dB(C) when the voice amplification system was turned on. The fundamental frequency (fo) of the teacher’s voice reduced by 1.4% with the use of the voice amplifier. The response of the students was overwhelmingly in support of the voice amplification system used during the recordings. 60-62% of female respondents and 68-79% of male respondents felt that the voice amplification system increased the audibility in the classroom.

    Conclusion: Leq [dB(C)] values from the mics around the classroom showed a considerable increase and thus the amplification system provides a good acoustic condition for the teacher to communicate with his students and a significant improvement in student’s comprehension of the teacher’s voice and learning conditions. There was a decrease in the fo of the teacher’s voice with the amplification, which is good for the teacher’s vocal health and reduces voice disorders while improving the quality and audibility of teacher’s voice. Majority of the students felt that the amplification system was beneficial during the teaching sessions and would like to see more amplifiers installed in more classrooms.

  • Jahn, Ilka
    et al.
    KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electric Power and Energy Systems.
    Hohn, Fabian
    KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electric Power and Energy Systems.
    Sharifabadi, Kamran
    Equinor ASA, Fornebu, Norway.
    Wang, Mian
    KU Leuven/EnergyVille, Leuven/Genk, Belgium.
    Chaffey, Geraint
    KU Leuven/EnergyVille, Leuven/Genk, Belgium.
    Norrga, Staffan
    KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electric Power and Energy Systems.
    Requirements for open specifications in multivendor HVDC protection systems2020Conference paper (Refereed)
    Abstract [en]

    Integrating multiterminal high-voltage direct-current (HVDC) connections into the existing alternating current (AC) power transmission system is a possible solution for transport of large amounts of renewable energy. Protection is considered a key enabler for multiterminal HVDC (MTDC) grids. Designing such a protection system is a challenge, in particular in a multivendor setup. Fault detection during transients might be required to achieve the stringent speed requirements for HVDC protection. This in turn requires knowledge about the expected system behaviour during faults and involves knowledge from all vendors supplying equipment in MTDC grid. Traditionally, HVDC projects are, however, supplied as turn-key solutions and the control and protection systems are the intellectual property of the vendors and not open for the HVDC end-users. This paper aims to provide a starting point on the discussion which information has to be shared between vendorsand HVDC end-users when designing MTDC grid protection. Simulations results show that missing information about certain aspects can lead to a failure of the protection system. A discussion on organization of the available information is added. Open specifications of the used equipment are useful to design safe and reliable MTDC grid protection.

  • Public defence: 2020-04-03 13:00 Join Zoom Meeting https://kth-se.zoom.us/j/664249709
    Garidis, Konstantinos
    KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electronics and Embedded systems.
    Applications of Si1-xGex alloys for Ge devices and monolithic 3D integration2020Doctoral thesis, comprehensive summary (Other academic)
    Abstract [en]

    As the semiconductor industry moves beyond the 10 nm node, power consumption constraints and reduction of the negative impact of parasitic elements become important. Silicon germanium (Si1−xGex) alloys have been used to amplify the performance of Si based devices and integrated circuits (ICs) for decades. Selective epitaxial growth of heavily doped Si and/or Si1−xGex is commonly employed to reduce the effect of parasitic resistance. Reducing the supply voltage leads to lower dynamic power consumption in complementary metal-oxide-semiconductor (CMOS) technology. Monolithic three-dimensional integration (M3D) is a technology that employs vertical stacking of the device tiers. This approach reduces the wiring length, effectively reducing interconnect delay, load capacitance and ultimately reducing the power consumption. Among the integration challenges M3D is facing, one can distinguish the available thermal budget for fabrication, the crystalline quality of the device active layer and finally the actual device or circuit performance.Germanium channel devices can benefit M3D integration. Germanium metal-oxide-semiconductor field-effect transistors (MOSFETs) can be fabricated at significantly lower temperatures than Si. In addition, they potentially can have higher performance compared to Si due to the superior electron and hole mobilities of Ge. Active layer transfer of crystalline quality layers is a key step in a M3D fabrication flow. Direct wafer bonding techniques offerthe possibility to transfer a Ge layer on a patterned wafer. This thesis studies the various applications of Si1−xGex films in M3D. An initial implementation of an in situ doped Si1−xGex film on silicon-on-insulator (SOI) and germanium substrates is first presented. A Si1−xGex film isgrown selectively on SOI substrates to be used as a contact electrode on Si nanowire biosensors. On Ge bulk substrates, in situdoped Si1−xGex is epitaxially grown to form p+-n junctions. The junction leakage current and the mechanisms at play are studied. The analysis ofthe junction performance provides insights on the junction leakage mechanisms,an important issue for the implementation of in situ doped Si1−xGex in M3D. A low temperature germanium-on-insulator (GOI) fabrication flow based on room temperature wafer bonding and etch back is presented in this work. The method suggested in the thesis produces high quality, crystalline Ge device layers with excellent uniformity. The thesis also reports on the development and integration of Si1−xGex in the GOI fabrication as an etch stop layer, enabling the stability of the layer transfer process. Finally this thesis presents Ge p-channel field-effect transistor (PFET) devices fabricated on the previously developed GOI substrates.The technologies presented in this thesis can be integrated in large scale Ge device fabrication. The low temperature GOI and Ge PFET fabrication methods are very well suited for sequential device fabrication. The processes and applications presented in this thesis meet the current thermal budget, device performance and active layer transfer demands for M3D technology.

  • Public defence: 2020-04-03 10:00 Stockholm
    Laakso, Miku
    KTH, School of Electrical Engineering and Computer Science (EECS), Intelligent systems, Micro and Nanosystems.
    Microfabrication and Integration Using Sub-Picosecond Laser Pulses and Magnetic Assembly2020Doctoral thesis, comprehensive summary (Other academic)
    Abstract [en]

    Microfabricated devices and systems have many exciting applications such as accelerometers for triggering the launching of airbags in cars, gyroscopes for sensing the rotations of mobile phones, and micromirror arrays for controlling light reflection in digital light projectors. These devices are currently produced using semiconductor manufacturing techniques, which are suitable for large volumes of mostly planar structures. However, they have limited economic viability for products with lower volumes, and they also constrain the three-dimensional (3D) structuring of the microdevices. Therefore, there is a need for new manufacturing techniques that are economically viable even for smaller volumes and allow truly 3D microdevice designs. To address this problem, this thesis presents developments in microfabrication and integration using two main methods: (1) The usage of sub-picosecond laser pulses for locally adding and modifying material and (2) the usage of an external magnetic field to handle fragile micrometric objects in order to assemble them into their target locations. These two methods are used for six main applications out of which four involve packaging and integrating microsystems, one involves the manufacturing of 3D microstructures, and one involves directly patterning microstructures on a surface.

    A key technology in the packaging and integration of microsystems, and a focus area of this thesis, is the manufacturing of through-substrate vias. They are used as electrical interconnections through device and package substrates. They allow smaller packages, which is a requirement, for example, for the Internet of Things where different types of microsensors and actuators are placed in our everyday environment. The first application related to the manufacturing of through-substrate vias is laser drilling of through-silicon holes. Laser drilling allows holes to be created where traditional etching methods might be uneconomical or unpractical. Laser drilling also allows the drilling of tilted holes, which can improve the radio-frequency performance of the vias. The second application is the magnetic assembly of metal conductors into holes in a glass substrate. Glass substrates have several benefits over silicon substrates, such as lower radio-frequency losses, but the production of through-glass vias is challenging due to the difficulty of creating regular holes through the glass. The magnetic assembly allows metal conductors to be placed into the holes in glass independent of the hole shape. This could lead to wider use of glass with its excellent properties as a packaging substrate for microsystems. The third application is through-substrate vias for high-temperature environments. These vias are manufactured by magnetically assembling metal conductors with low thermal expansion into holes in a silicon substrate. The low thermal expansion leads to reduced stresses at elevated temperatures. This could allow using through-substrate vias to reduce package sizes even in demanding high-temperature environments found, for example, in the space industry.

    The fourth and last application related to the packaging and integrating microsystems is the vertical assembly of microchips using an external magnetic field. Microsystem fabrication is focused on in-plane structures, but some applications require or would benefit from out-of-plane structures. Examples of such applications are a biosensor placed inside a microneedle inserted into tissue or flow sensors bending in the flow. Manufacturing the out-of-plane structures on the same substrate with other structures requires complicated manufacturing techniques and occupies a large surface area. When using the vertical assembly process, the out-of-plane structures can be manufactured on a separate substrate using standard microfabrication techniques, and the out-of-plane structures can then be assembled afterward in a vertical orientation on a receiving substrate.

    Manufacturing of 3D microstructures is not trivial using the standard micromanufacturing techniques. Free-form 3D printing of submicrometric features is possible using two-photon polymerization, but the material properties of polymers are not comparable to those of silica glass. This thesis demonstrates 3D printing of silica glass with submicrometric features using sub-picosecond laser pulses. This new 3D freedom in micromanufacturing could be used, for example, in building more complicated micro-opto-electro-mechanical systems.

    Directly patterning microstructures on a surface is possible by exposing the surface to laser pulses. These structures can affect the optical and wetting properties of the surfaces. More specifically, periodic ripple structures can act as diffraction gratings, altering the optical reflection properties of the surface. Exposure to sub-picosecond laser pulses can also cause chemical changes on the surface, and these changes can potentially affect the reflection properties. This thesis demonstrates that the chemical changes indeed affect the reflection properties, and this information could be used when manufacturing ripple patterns, for example, for security markings or for decorative use.

  • Losenius Lindahl, Viktoria
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH).
    Evaluating Laboratory Information Systems to Ensure Patient Safety and Analytical Quality in Clinical Laboratory Processes2018Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Laboratory information systems (LIS) are an established IT-support system currently used by clinical laboratories. Traditional methods in clinical laboratories are improved through LIS and automation. Such changes generate more efficient and productive clinical laboratories, allowing faster diagnosis. This study evaluates how a LIS can increase efficiency, ensure patient safety, and improve analytical quality in the laboratory processes. A literature review was made as first-step and used to generate deeper knowledge about the origin of LIS and traditional methods within the clinical laboratories. A qualitative, interview-based investigation with eight interviewees primarily from clinical laboratories was used to evaluate LIS. The study showed that an LIS must support laboratory workflow to ensure patient safety and analytical quality management. During the creation of LIS, present and expected national guidelines must be considered. Economically and socially sustainable LIS should be considered in the transition of an updated or new LIS. Further studies should investigate all clinical laboratories and vendors. 

  • Sivard, Gunilla
    KTH, School of Industrial Engineering and Management (ITM), Production Engineering.
    AN INTEGRATED ARCHITECTURE FOR FUNCTIONAL PRODUCTS2003In: INTERNATIONAL CONFERENCE ON ENGINEERING DESIGNICED 03 STOCKHOLM, AUGUST 19-21, 2003 / [ed] Folkeson, A.; Gralen, K.; Norell, M.; Sellgren, U, 2003Conference paper (Refereed)
    Abstract [en]

    There is an increasing trend in industry of selling services in addition to physical products. Infunctional sales, companies sell the service of delivering the function of the product asopposed to selling the product itself. While there is a long tradition of structured developmentand management of purely physical goods, there is a lack of such methods for services, andfor the mixture of service and goods - so called functional products.This paper describes an effort of defining a formal model of services, contributing to theefforts of structuring, visualizing and managing functional products in analogue with purelyphysical products.A modular architecture for services is presented, based on the paradigm of defining servicesas discrete, decomposable activities. Sub activities, so called service modules, are described interms of the activity’s properties and interfaces to other service modules. Since the method isbased on principles for uncoupled design, it provides support for creating transparent andflexible architectures with a minimum of built in dependencies between functions and theirphysical implementation.The service module aggregation and representation principles are described in relation to anindustrial distribution case.

  • Mokhberi, Shiva
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH).
    Evaluation of a Sensor-Based System for Ergonomic Risk Assessment among Hairdressing Students2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Occupational upper extremity disorders have become a major issue in modern society. Poorly designed workplaces, high job demands, and incorrect work-habits can lead to the development of upper extremity disorders (UEDs) in the workplace. This issue not only causes health-relatedproblems for the individual but also forms a significant economic burden on society due to sickleaves, healthcare and untimely exit of affected individuals from the workforce.

    The risk of developing occupational UEDs varies with different professions. The European Agency for Safety and Health at Work (EU-OSHA) has recognized significant occupational health risks associated with the hairdressing profession. It has been estimated that UEDs are five times more prevalent amongst hairdressers than other professions.

    Qualitative risk assessment tools based on self-reports and observation have been used to identify the risks of developing UEDs with hairdressing profession before. However, a quantitative risk assessment tool that provides objective data on work posture is more precise and objective than self-report and observation. This data can help to identify the risks of developing UEDs associated with each hairdressing task. Furthermore, it can enable self-assessment of workload and posture awareness by providing feedback to the user.

    Inertial Measurement Units (IMU) as part of a wearable system developed at KTH were used in this study to investigate the risks of developing UEDs for hairdressing students. The feasibility of using a feedback function for providing posture awareness was also evaluated by comparing the measurements obtained with and without using the feedback function. Twelve hairdresser students were enrolled in the study.

    The percentage of time for elevated angles above 30°, 60° and 90° for arms, and above 45° or less than 0° for the trunk flexion is presented. In addition, 10th and 90th percentiles (°) of arms and trunk angular distribution is presented. The result of a statistical analysis performed on data with and without feedback was used to evaluate the effectivity of using the feedback function in preventing the development of occupational UEDs. A System Usability Scale (SUS) questionnaire was used to evaluate the overall usability of the system.

    The result of this study confirms that the hairdressing profession falls in the high-risk category for developing UEDs. The use of this technical system has enabled a precise risk assessment evaluation of each hairdressing task. Such data can be used as a foundation for improving the ergonomic design of the workplace. The feasibility of using the feedback function as a prevention tool on the individual level is highly dependent on the individuals’ motivation and their attitude towards changing their work habits. However, the results in general, indicate a decrease in the abduction angle (°) for both left and right arm when the feedback function is used. For example, the 90th percentile abduction angles (°) for left arm (all 12 subjects) during the drying part of one fundamental work-cycle decreased from a value of 60.4° to 58.2° when the feedback function was used. The 90th percentile abduction angles (°) for the right arm during the same part of the fundamental work-cycle decreased from an angle of 53.1° to 51.4°.

    The SUS score of 75.6 indicates good overall usability for the system.

  • Zhang, Xiran
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics.
    3D Printed Soft Robot Gripper with Closed-Loop Control2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    This project aims to build up a soft robotic gripper that mimics human hands and design a closedloop control system. A soft gripper model is established with Finite Element Method (FEM) to describe the relation between air pressure input and gripper deformation. The best soft gripper dimensions are selected according to the FEM model and the gripper is then fabricated with Fused Deposition Modeling (FDM) 3D printing method. Closed-loop angle control before the soft gripper touches the object is used to ensure a precise grasp. A camera sensor is used for the acquisition of the bending angle and a pressure regulator is applied to supply the air pressure. A closed-loop experiment platform is built based on a proportional-integral (PI) controller to realize the precise deformation control of the soft gripper. Finally, the grasp of some soft or brittle objects using the soft gripper is performed as a demonstration.

  • Valle Olivera, Nicole
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH).
    Modeling and Evaluation of a Finite Element Cervical Spinal Cord for Injury Assessment2020Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Motor vehicles collisions and falls have gradually increase the risk for spinal cord injuries. An increased knowledge of the spinal behavior and its injury mechanisms can be used as preventive strategies. Total Human Model for Safety (THUMS) SAFER is used as a tool for injury prevention, however, there is a lack of studies that evaluate the spinal cord injuries. The aim of this thesis is to implement a cervical spinal cord into the THUMS model.

    The mesh element quality was modified and the spinal cord was further adjusted for a correct insertion into the THUMS. The strain of the posterior and anterior surface of the cervical spinal cord during a head flexion were analyzed against experiments. Subsequently, a comparison of the head kinematics in frontal collision of the THUMS with and without the cervical spinal cord was performed.

    A refinement of the mesh element quality for a suitable computational time was achieved. The strain evaluation of the the spinal cord showed the same behavior as in the experiment for the posterior surface but the results were contradictory for the anterior surface. The results of the head kinematics with and without spinal cord showed no good correlation with the experimental data. Moreover, the models exhibited a bigger difference between them during the extension of the head than flexion.

    A further improvement of the mesh element quality required smaller element size. Nonetheless, it is important to consider that computational time increases with a decrease of element size. Several factors were critical for the strain comparison, such as the lack of information for the calculation of the strain. The difference in head kinematics from the experiment may be due to the material properties of the neck skin and the lack of the active muscles. Moreover, the contact constraints in the model may result in the differences between the THUMS models.

    In general, the spinal cord has been refined to obtain a favorable computational time. The evaluations have indicated that further modifications in the neck skin and contact constraints are needed for a better resemblance with the human body. Likewise, further validations against experimental studies are suggested.

  • Guazzo, Alessandro
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH).
    Deep Learning for PET Imaging: From Denoising to Learned Primal-Dual Reconstruction2020Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    PET imaging is a key tool in the fight against cancer. One of the main issues of PET imaging is the high level of noise that characterizes the reconstructed image, during this project we implemented several algorithms with the aim of improving the reconstruction of PET images exploiting the power of Neural Networks. First, we developed a simple denoiser that improves the quality of an image that has already been reconstructed with a reconstruction algorithm like the Maximum Likelihood Expectation Maximization. Then we implemented two Neural Network based iterative reconstruction algorithms that reconstruct directly an image starting from the measured data rearranged into sinograms, thus removing the dependence of the reconstruction result from the initial reconstruction needed by the denoiser. Finally, we used the most promising approach, among the developed ones, to reconstruct images from data acquired with the KTH MTH microCT - miniPET.

  • Högsäter Myhr, Mattias
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.).
    Techniques for Improving Weldability and Testing of Weld-bonded Joints2020Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Increased understanding of improving the weld-bonding process is essential in order for the automotive industry to constantly develop their designs and production processes for the energy-efficient vehicles of the future. The joining method of weld-bonding, a hybrid method of adhesive joining and resistance spot welding, has introduced new challenges in the manufacturing and design of auto bodies. At the same time as adhesives are more incorporated in the joint configurations’, continuous development and implementation of ultra-high-strength steels, the complexity of the weld-bonding process increases. This thesis presents new approaches of parameter set-up and testing methods for weld-bonded joints. Aiming to displace the adhesive from the welding zone new parameters are tested by producing 1-D weld lobes that are analyzed in a screening matrix to find the factor with greatest effect on the weldability. Laser displacement measurement is a new method used to measure the amount of adhesive present in the joint during the welding cycle. The amount of adhesive plays an important role in the process and the intention is to get an assessment of how different thicknesses affect the weldability. The results showed that the different parameters have different influential behavior depending on the sheet thickness of the welded material. From the main effect screening, cap type has the greatest influence on the weld current range followed by the addition of an extra pre-pulse. The measured amount of adhesive present immediately before the weld pulse is very small. An interesting observation is that samples showing the largest increase in weld current range, compared to the corresponding reference were the samples which had most adhesive present in the contact area. The study showed that there are new parameters that improve the weldability of a weld-bonded joint that can be utilized using existing welding equipment. Finally, the study has shown that many possibilities exists to continue exploit within the field of weld-bonding, to be able to further utilize the effectiveness of weld-bonded joints in future lightweight autobody design.

  • Ouadria, Michel Sebastian
    et al.
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems.
    Ciobanu, Ann-Stephanie
    KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems.
    Sentimentanalys av svenskt aktieforum för att förutspå aktierörelse2020Independent thesis Basic level (professional degree), 10 credits / 15 HE creditsStudent thesis
    Abstract [en]

    The present study examines the possibility of predicting stock movement on a daily basis with sentiment analysis of posts in a swedish stock trading forum. Sentiment analysis is used to find subjectivity in the form of emotions (sentiment) from text. Textdata was extracted from a stock forum to predict the share movement of the related share. All data was aggregated within a fixed period of two years. The analysis utilizes machine learning to train three machine learning models with textdata and stockdata. The result showed no clear correlation between sentiment and stock movement. Furthermore, the result was not able to replicate accuracy as previous work in the field. The highest accuracy achieved with the models was calculated at 64%.

  • Shenoy, Prithvi
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics.
    A Vehicular Ad Hoc Network Based Localization for a City Bus2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    City busses are operated on roads where the GPS signal is weak, because of the tall buildings surrounding these roads. The localization of city busses, needs to therefore rely on alternate technique in order to improve the accuracy. Recent standardization of inter vehicular communication has made this a readily available tool which can be used for localization. This thesis presents an approach towards localization of a city bus by means of vehicular ad hoc network. The two main components of localization by this approach is the initialization of location estimate component, and the real time location estimation component. In particular, the thesis develops the use of minimum mean square estimation for initialization and an extended Kalman filtering approach for real time location estimation. The localization method is mathematically described, considering the operating scenarios of a city bus. The accuracy of the proposed method is mathematically evaluated. The developed localization method is implemented in a simulation tool kit for inter vehicular communication. Simulation experiments were performed for operating scenarios of city bus. The result of initialization by minimum mean square error is compared to that of initialization by GPS, in-terms of localization accuracy. Different setups of road side units are compared in-terms of accuracy and update interval. The results show that the proposed method is feasible for localization of a city bus. This thesis was carried out in association with Scania AB, Södertälje.

  • Törngren, Martin
    et al.
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics. KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Machine Design (Div.). KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Embedded Control Systems. KTH, Superseded Departments (pre-2005), Machine Design.
    Asplund, Fredrik
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics. KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Machine Design (Div.). KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Embedded Control Systems.
    Ericson, Tor
    ÅF Digital Solutions AB.
    Granbom, Catrin
    Ericsson AB.
    Herzog, Erik
    Saab Aeronautics.
    Lu, Zhonghai
    KTH, Superseded Departments (pre-2005), Electronic Systems Design. KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electronics and Embedded systems, Electronic and embedded systems.
    Magnusson, Mats
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Integrated Product Development. KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Product Innovation Technology.
    Månsson, Maria
    Prevas AB.
    Norrwing, Stefan
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics.
    Olsson, Johanna
    KTH, School of Industrial Engineering and Management (ITM), Machine Design (Dept.), Mechatronics.
    Öberg, Johnny
    KTH, Superseded Departments (pre-2005), Microelectronics and Information Technology, IMIT. KTH, Superseded Departments (pre-2005), Electronic Systems Design. KTH, Superseded Departments (pre-2005), Electrical Systems. KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electronics and Embedded systems, Electronic and embedded systems.
    Competence Networks in the Era of CPS: Lessons Learnt in the ICES Cross-Disciplinary and Multi-domain Center2019In: Proceedings of the International Workshop on Design, Modeling, and Evaluation of Cyber Physical Systems, CyPhy 2019: Workshop on Embedded Systems and Cyber-Physical Systems Education / [ed] Chamberlain R., Edin Grimheden M., Taha W., Cham, 2019, Vol. 11971, p. 264-283Conference paper (Refereed)
    Abstract [en]

    Cyber-Physical Systems (CPS) are evolving to become more intelligent, autonomous and collaborating, playing an important role in societal infrastructure. The amount of knowledge required in developing and managing future CPS will be unprecedented, leading to stronger needs for collaboration, competence provisioning, continuous learning and renewal of education. This is where “competence” (or learning) “networks” involving academia and industry play an important role. We elaborate and discuss needs, lessons learnt and challenges for such competence networks in the context of CPS. We draw upon our experiences gained from ICES - the KTH-industry cross-disciplinary and multi-domain competence network which in 2019 has been operational for 11 years, growing from 6 to more than 30 participating organizations. The ICES network focuses on activities to support students, industrial engineers and managers, and academic faculty, acting as a network, catalyst and competence provider directed towards these stakeholders. We elaborate challenges faced during the operation of ICES including the lack of prioritization of competence networks and education, the paradox with strong needs for competence networks but perceived lack of time, the challenges of reaching out to stakeholders, and fragmented efforts addressing competence provisioning. We finally discuss ways forward. In conclusion, we believe that the ICES type of network could be relevant in many other areas characterized by complex systems.

  • Millinger, Jonas
    KTH, School of Electrical Engineering and Computer Science (EECS), Electrical Engineering, Electric Power and Energy Systems.
    Slotless High-Performance Electric Drives: Harmonic Loss Analysis, Sensorless Control, and Inverter Design2020Doctoral thesis, comprehensive summary (Other academic)
    Abstract [en]

    Electric motor drives are a cornerstone for many fundamental functions in today’s society, and their expected dominance within the transport sector is considered one of the main keys to abate global warming. Subsequently, the incentives for improved performance of electric motor drives cannot be sufficiently emphasized. This thesis strives for improved performance in electric motor drives of industrial power tools, such as the nutrunner system. A typical nutrunner system comprises a wall-mount power converter and a cable-connected, lightweight (handheld) nutrunner. The converter feeds the electric motor with a pulse-width modulated signal, which, apart from the desired fundamental component, also contains undesired high-frequency harmonics. As a consequence, harmonic losses are generated in the motor which reduce the performance. Harmonic losses occur in every conductive part of the motor, but rotor losses are considered more challenging due to the poor heat transfer across the air gap. Excessive rotor temperatures can cause premature bearing failure and even irreversible demagnetization of the permanent-magnets. The conventional solution is to use a large inductive motor filter to suppress the harmonic currents fed to the motor. However, the recent emerge of wide-bandgap transistors enables significantlyhigher switching frequencies in electric motor drives compared to their conventional, silicon-based counterparts, which can reduce the size of the required motor filter. The aim of this work is to eliminate the wall-mountconverter and make it sufficiently small for integration inside the nutrunner unit. Optimization of electric motor drives using wide-bandgap technology requires accurate models for the prediction of harmonic phenomena in electric machinery. The main focus of the thesis is to develop an accurate model for the predictionof harmonic losses in slotless permanent-magnet motors when fed by a wide-bandgap inverter. As a first step, the small-signal magnetic behaviors (including power losses) of silicon and carbon steel are characterized. Next, a computationally efficient, three-dimensional finite-element model for simulation of harmonic motor losses, is developed. The model is experimentally validated using a broad range of rotor-magnet segment thicknesses. The developed models show that wide-bandgap technology can effectively contribute to the elimination of the inductive motor filter. In the second part of the thesis, a novel sensorless control-method (enabled by a filter-less motor drive topology) is developed, and experimentally evaluated. The results show an estimation error of the rotor position below 2 degrees, which, in contrast to conventional slotted machinery, is practically unaffected by the load (current) level. In the third part of the thesis, a compact, filter-less wide-bandgap inverter is developed and experimentally evaluated. Despite operation at significantly higher switching frequencies, inverter power losses can be halved, which further reduces the system weight (due to a smaller heat-sink). To summarize, the utilization of wide-bandgap transistors enables substantial improvements in terms of reduced system weight, complexity, and power losses in slotless permanent-magnet motor drives.

  • Janik, Adrianna
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Interpretability of a Deep Learning Model for Semantic Segmentation: Example of Remote Sensing Application2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Understanding a black-box model is a major problem in domains that relies on model predictions in critical tasks. If solved, can help to evaluate the trustworthiness of a model. This thesis proposes a user-centric approach to black-box interpretability. It addresses the problem in semantic segmentation setting with an example of humanitarian remote sensing application for building detection.

    The question that drives this work was, Can existing methods for explaining black-box classifiers be used for a deep learning semantic segmentation model? We approached this problem with exploratory qualitative research involving a case study and human evaluation.

    The study showed that it is possible to explain a segmentation model with adapted methods for classifiers but not without a cost. The specificity of the model is likely to be lost in the process. The sole process could include introducing artificial classes or fragmenting image into super-pixels.

    Other approaches are necessary to mitigate identified drawback. The main contribution of this work is an interactive visualisation approach for exploring learned latent space via a deep segmenter, named U-Net, evaluated with a user study involving 45 respondents. We developed an artefact (accessible online) to evaluate the approach with the survey. It presents an example of this approach with a real-world satellite image dataset.

    In the evaluation study, the majority of users had a computer science background (80%), including a large percentage of users with machine learning specialisation (44.4% of all respondents). The model distinguishes rurality vs urbanization (58% of users). External quantitative comparison of building densities of each city concerning the location in the latent space confirmed the later. The representation of the model was found faithful to the underlying model (62% of users).

    Preliminary results show the utility of the pursued approach in the application domain. Limited possibility to present complex model visually requires further investigation.

  • Zareafifi, Farhad
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Link Criticality Characterization for Network Optimization: An approach to reduce packet loss rate in packet-switched networks2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Network technologies are continuously advancing and attracting ever-growing interests from the industry and society. Network users expect better experience and performance every day. Consequently, network operators need to improve the quality of their services. One way to achieve this goal entails over-provisioning the network resources, which is not economically efficient as it imposes unnecessary costs. Another way is to employ Traffic Engineering (TE) solutions to optimally utilize the current underlying resources by managing traffic distribution in the network. In this thesis, we consider packet-switched Networks (PSN), which allows messages to be split across multiple packets as in today’s Internet. Traffic engineering in PSN is a well-known topic yet current solutions fail to make efficient utilization of the network resources.

    The goal of the TE process is to compute a traffic distribution in the network that optimizes a given objective function while satisfying the network capacity constraints (e.g., do not overflow the link capacity with an excessive amount of traffic). A critical aspect of TE tools is the ability to capture the impact of routing a certain amount of traffic through a certain link, also referred as the link criticality function. Today’s TE tools rely on simplistic link criticality functions that are inaccurate in capturing the network-wide performance of the computed traffic distribution. A good link criticality function allows the TE tools to distribute the traffic in a way that it achieves close-to-optimal network performance, e.g., in terms of packet loss and possibly packet latencies. In this thesis, we embark upon the study of link criticality functions and introduce four different criticality functions called: 1) LeakyCap, 2) LeakyReLU, 3) SoftCap, and 4) Softplus. We compare and evaluate these four functions with the traditional link criticality function defined by Fortz and Thorup, which aims at capturing the performance degradation of a link given its utilization.

    To assess the proposed link criticality functions, we designed 57 network scenarios and showed how the link criticality functions affect network performance in terms of packet loss. We used different topologies and considered both constant and bursty types of traffic. Based on our results, the most reliable and effective link criticality function for determining traffic distribution rates is Softplus. Softplus outperformed Fortz function in 79% of experiments and was comparable in the remaining 21% of the cases.

  • Baccelli, Alessio
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Knowledge distillation: a new approach towards LSTM interpretability2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Nowadays, machine learning is been used in a different ranges of domains, from finance to predictive maintenance, from image recognition to natural language processing. In particular, the rise of new techniques and new architectures have allowed neural network models to reach in many cases human-level performance in different tasks and domains. However, this boost in performance often comes with a lack of interpretability, which is crucial in some particular domains, such as the one of medicine. Many studies have therefore been conducted with a particular focus on improving interpretability of neural networks models. This thesis project aims to apply knowledge-distillation, a technique which has been successfully applied to some kind of neural networks, to a LSTM, in such a way to simplify the model and obtain a CART tree. The underlying hypothesis is that this kind of tree is more interpretable than the original LSTM, as well as more accurate with respect to a standard decision tree. The dataset that has been used is the Sleep Data dataset, with time series of different biological signals coming from different patients. In particular, the network has been trained to recognize sleep apnea episodes. The CART trained with the help of the LSTM is actually performing slightly better with respect to the baseline. However, both the LSTM and the CART are not able to reach state-of-the-art performance. Therefore, further researches and improvements are needed in order to obtain better and more meaningful results.

  • Maqsood, Bilal
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Implementation and performance analysis of software defined radio (SDR) based LTE platform for truck connectivity application2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    In today’s modern era of technology mobile communication is evolving with a higher pace than ever before. New features and applications are added in the existing networks each day. The faster development pace requires a faster way to prototype and test the mobile communication standards/ applications to offer faster delivery to the end user. In traditional practices hardware updates and new features growth take long time to market implementations. Technology tends to be obsolete by the time it is to be launched to the market. The reason being long time required for hardware production. However, in the recent days the trend is changing with the emergence of open source cellular stacks to be used with affordable software defined radio (SDR) hardware platforms.

    Long term evolution (LTE) open source cellular stacks along with the SDR technology are widely used in research these days. However, the performance and limitations of these SDR based open source cellular stacks needs to be explored. In this project a thorough study is performed to access the performance of an open source SDR based LTE user equipment (UE) software stack. A prototype of Category 4 LTE modem is implemented using the srsLTE application suite. Performance analysis is done by looking into the datarate, SNR and radio frequency (RF) characteristics of the implemented solution for multiple system bandwidth settings. A performance comparison is presented between the high performance SDR platform Universal Software Radio Peripheral x310 and the LimeSDR. The results show that the SDR technology is capable of handling wideband signals like LTE. The choice of SDR hardware platform and open source cellular stack depends on the application. The chosen solution for this project i.e. srsLTE performed well for LTE bandwidths 10 MHz and above in terms of downlink data rate. However, the radio frequency characteristics of selected SDR platforms do not comply fully with the 3GPP standard requirements.

  • Vasan, Srinath
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Truck Connectivity Platform Using Software Defined Radios2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Hardware updates and feature updates of electronic equipment can take a long time for market implementations. Technology tends to become obsolete by the time the update is launched into the market. In the automotive industry, the service entertainment module (SEM) containing the core logic for Amplitude Modulation (AM), Frequency Modulation (FM), and Digital Audio Broadcasting (DAB) radio tuners lack the facility for software-based updates. One of the major challenges for the infotainment/entertainment modules in the automotive industry is to have a single radio module that can support software updates and multistandard radio technologies from different countries. Software Defined Radios (SDR) can be used to deal with this problem as SDR is a platform to prototype or develop new communication protocols as well as add new features or capabilities to the existing protocols without requiring major capital expenditures. SDR is a cost-effective radio platform because it can update radio equipment on the fly and provide additional functionality without requiring hardware modifications. By means of software instead of hardware updates, development loops could be shortened and manufacturing costs are reduced. In this thesis, a thorough comparison of a hardware-based tuner and a software-based tuner is performed in the presence of AM/FM/DAB modulated signals based on the metrics specified in ETSI EN 303 345-1 including sensitivity, adjacent channel suppression, far-off channel suppression, cross-modulation, third-order intermodulation, etc. After performance evaluation and comparison, it can be proven that the SDR based system can perform as well as the hardware tuner used in the SEM unit.

  • Abdirahman Adami, Adnan
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Actors Cooperation Analysis: A Techo-economic Study on Smart City Paradigm2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Modern cities must overcome complex challenges to achieve socio-economic development and to improve the quality of life as the urban population is rapidly increasing. The concept of smart citie s is a response to these challenges. Thus, emerging technologies that are key enablers for the development of a smart city are said to be IoT and 5G. To deploy such technologies , however, may be expensive and requires the involvement of multiple actors. Hence, lack of cooperation and coordination for planning, financing, deploying and managing the city’s operational networks makes it even more difficult to overcome such challenges. Further, waste management companies and parking services operators in a city have expensive operation costs and services inefficiency due to little utilization of IoT-based solutions. This paper identifies and analyzes smart city ecosyst e ms, value networks, actors, actor’s roles, and business models in order to illustrate business relationships and provide business opportunities in the development of smart and sustainable cities through cooperation and collaboration among involved actors . Target actors that this study focuse s are on Mobile Network Operators, Parking Services Operators, and Waste Management Companies, and uses smart parking and smart waste collection as use-cases. Results show several cooperative business scenarios that can lead to successful business relationships and opportunities.

  • Kimblad, Jacob
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Extending Timing Analysis for Non­ Preemptive Task Sets on Multicore Under the AER Model2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Real-time multi-core systems with shared memory are harder to analyze due to varying execution times caused by congestion in accessing the shared memory. A promising way to make these systems more deterministic is the Acquisition-Execution-Restitution (AER)-model which takes a memory-centric approach to scheduling which enables the multicore problem to be seen as a single core scheduling problem. Memorycentric scheduling means that instead of scheduling tasks onto cores the memory accesses are scheduled onto the memory, where in this case the memory phases are allowed to access the shared memory nonpreemptively during their allocated slots. In this thesis an extended version of the AER-model and a new Idle-time Insertion Policy (IIP), which can make decisions on whether to leave resources idle when there is work to be done as an attempt to improve schedulability, is presented. These can be used in conjunction with a state-of-the-art scheduling analysis tool to perform timing analysis on job sets under the AER-model. The extended AER-model introduces scheduling windows for the Acquisition (A)-phase and Restitution (R)-phase of tasks in a way that no further precedence constraints need to be considered by the scheduler as not to schedule any of the A-phase or R-phase in the wrong order. The developed IIP is then used by the analysis tool to make sure that there is at least one free core in the system before a new task is scheduled as this needs to be put into consideration in a memory-centric scheduler. The extended AER-model and the developed IIP are then benchmarked with task sets coherent to the AUTOSAR-framework. The benchmarks explore the schedulability ratio when varying task amount, core amount, scheduling algorithms and the window ratio for the phases. The results show that Earliest Deadline First (EDF) outperforms other algorithms that are special versions of Rate-Monotonic Scheduling (RMS) where priorities are assigned based on a combination of a tasks period and its utilization. The results also show that tasks with a window ratio of 60% as well as task sets of sizes between 100 to 150 yields the highest schedulability ratio.

  • Tang, Guanqian
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Monocular Depth Prediction in Deep Neural Networks2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    With the development of artificial neural network (ANN), it has been introduced in more and more computer vision tasks. Convolutional neural networks (CNNs) are widely used in object detection, object tracking, and semantic segmentation, achieving great performance improvement than traditional algorithms. As a classical topic in computer vision, the exploration of applying deep CNNs for depth recovery from monocular images is popular, since the single-view image is more common than stereo image pair and video. However, due to the lack of motion and geometry information, monocular depth estimation is much more difficult.

    This thesis aims at investigating depth prediction from single images by exploiting state-of-the-art deep CNN models. Two neural networks are studied: the first network uses the idea of a global and local network, and the other one adopts a deeper fully convolutional network by using a pre-trained backbone CNN (ResNet or DenseNet). We compare the performance of the two networks and the result shows that the deeper convolutional neural network with the pre-trained backbone can achieve better performance. The pre-trained model can significantly accelerate the training process. We also find that the amount of training dataset is essential for CNN-based monocular depth prediction.

  • Tomaszuk, Michal
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Remote Control of Forest Machinery Using WiFi2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Automation of the forest industry has for over 30 years been an important subject of research, which could reduce the human workload and costs significantly. However, there are still many problems to be solved, such as enabling the communication between the heavy machinery in a forest and a remote base. High speed and reliable communication is the key to automated operations and remote control of machinery. This thesis investigates the feasibility and performance of IEEE 802.11n/ac WiFi hardware to provide high-bandwidth connection in a forest. In this project, the propagation of WiFi signals in the2.4 GHz and 5 GHz bands in a typical Nordic forest environment has been simulated using specialized radio propagation software employing ray-tracing and different diffraction models to evaluate the path loss and signal strength. The simulations show that the idea is feasible if high-gain directional antennas are employed, as connections of sufficiently high speed (400+ Mbps for the 5 GHz band) can potentially be established for typical working distances,i.e. 300m. We then designed a directional antenna system and evaluated it in a real Nordic forest environment. We found that by manually aligning the antennas in a forest, reliable connections could be achieved up to 50 m without line-of-sight, however higher distances result in significantly lower speeds (13.3 Mbps at 80 m and 1.21 Mbps at 100 m) due to antenna misalignment. It is however possible to construct a more accurate, automated alignment system, which could replicate the simulation results and fully solve the problem of communication.

  • Shao, Jingnan
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Characterization of FPGA-based Arbiter Physical Unclonable Functions2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    The security of service, confidential data, and intellectual property are threatened by physical attacks, which usually include reading and tampering the data. In many cases, attackers can have access to the tools and equipment that can be used to read the memory or corrupt it, either by invasive or non-invasive means. The secret keys used by cryptographic algorithms are usually stored in a memory. Physical unclonable functions (PUFs) are promising to deal with such vulnerabilities since, in the case of PUFs, the keys are generated only when required and do not need to be stored on a powered-off chip. PUFs use the inherent variations in the manufacturing process to generate chip-unique output sequences (response) to a query (challenge). These variations are random, device-unique, hard to replicate even by the same manufacturer using identical process, equipment and settings, and supposed to be static, making the PUF an ideal candidate for generation of cryptographic keys.

    This thesis work focuses on a delay-based PUF called arbiter PUF. It utilizes the intrinsic propagation delay differences of two symmetrical paths. In this work, an arbiter PUF implemented in Altera FPGA has been evaluated. The implementation includes Verilog HDL coding, placement and routing, and the communication methods between PC and FPGAs to make testing more efficient.

    The experimental results were analyzed based on three criteria, reliability, uniqueness, and uniformity. Experimental results show that the arbiter PUF is reliable with respect to temperature variations, although the bit error rate increases as the temperature difference becomes larger. Results also reveal that the uniqueness of the PUFs on each FPGA device is particularly low but on the other hand, the proportions of different response bits are uniform after symmetric routing is performed.

  • Gebretsadkan Kidane, Netsanet
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Hudi on Hops: Incremental Processing and Fast Data Ingestion for Hops2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    In the era of big data, data is flooding from numerous data sources and many companies have been utilizing different types of tools to load and process data from various sources in a data lake. The major challenges where different companies are facing these days are how to update data into an existing dataset without having to read the entire dataset and overwriting it to accommodate the changes which have a negative impact on the performance. Besides this, finding a way to capture and track changed data in a big data lake as the system gets complex with large amounts of data to maintain and query is another challenge. Web platforms such as Hopsworks are also facing these problems without having an efficient mechanism to modify an existing processed results and pull out only changed data which could be useful to meet the processing needs of an organization.

    The challenge of accommodating row level changes in an efficient and effective manner is solved by integrating Hudi with Hops. This takes advantage of Hudi’s upsert mechanism which uses Bloom indexing to significantly speed up the ability of looking up records across partitions. Hudi indexing maps a record key into the file id without scanning over every record in the dataset. In addition, each successful data ingestion is stored in Apache Hudi format stamped with commit timeline. This commit timeline is needed for the incremental processing mainly to pull updated rows since a specified instant of time and obtain change logs from a dataset. Hence, incremental pulls are realized through the monotonically increasing commit time line. Similarly, incremental updates are realized over a time column (key expression) that allows Hudi to update rows based on this time column.

    HoodieDeltaStreamer utility and DataSource API are used for the integration of Hudi with Hops and Feature store. As a result, this provided a fabulous way of ingesting and extracting row level updates where its performance can further be enhanced by the configurations of the shuffle parallelism and other spark parameter configurations since Hudi is a spark based library.

  • Laaksonen, Oskar
    KTH, School of Industrial Engineering and Management (ITM), Learning.
    Kemins tre perspektiv i undervisningenmakroskopiskt, symboliskt och submikroskopiskt: En tematisk analys baserad på semistrukturerade intervjuer2020Independent thesis Advanced level (professional degree), 10 credits / 15 HE creditsStudent thesis
    Abstract [en]

    In this essay it is described in what way three perspectives (macroscopic,submicroscopic, symbolic) of chemistry occur in teachers’ chemistryeducation.Data was collected using semi structured interviews with 13 teachers. Theinterviews were recorded and transcribed. Some of the teachers work withchemistry education in the upper stage of the primary school and some workin upper secondary education (high school). Then thematic analysis wasapplied on the transcribed interviews.The following themes were found in the resulting analysis:

    • Appropriate substance

    • The student’s mental picture

    • Unaware awareness

    • Two perspectives at a time

    In the theme appropriate substance it was found that teachers often usesubstances in their education that are already known to the student.In the theme the student’s mental picture it was found that teachersgenerally try to help the student imagine how chemistry works.In the theme unaware awareness it was found that the three perspectivesoccurred in the teachers’ education, but the words were only applied by one ofthe teachers systematically. In the other teachers’ teaching the concepts werepart of the education even if the teachers weren’t aware of the terminology.In the theme two perspectives at a time it was found that the teachersusually applied two perspectives but seldom all three at the same time.The three perspectives (the concepts) exist in all teacher’s education but mostof the teachers use other expressions. The concepts are used together andunstructured which may contribute students’ misconceptions. Only oneteacher educates systematically using the terminology macroscopic,submicrosopic and symbolic.

  • Barrios, Karolina
    et al.
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Tudose Fuentes, Amanda
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Social manipulation och phishing: Vilka brister finns i dagens skydd och hur kan de förbättras?2019Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
    Abstract [en]

    Companies in todays society must protect themselves against different types of cyber attacks. Since the attacks come in different forms, phishing attacks are one of the most common security issues that both individuals and companies face in keeping their information secure. Hackers are using email, social media, phone calls, SMS and any form of communication they can to steal valuable data like passwords, credit card numbers, or other sensitive information. Companies are particularly a target and it is estimated that cyberattacks cost the global economy 500 billion dollars per year, where phishing stands for 90% of that number.

    Today, everything is put online and the safety of personal credentials is at risk. Phishing can be seen as one of the oldest and easiest ways of stealing information from people. It also has a simple approach as the attacker sends an email to a victim, and the victim enters a fake website and gives the attacker personal information. All without realizing it.

    This report attempts to identify the deficiencies found in the protections used against phishing and how they can be improved to prevent a phishing attack. They also try to look for any protection or combination of protection that provides better protection. The reports divides the solution in different categories - machine learning, network solutions, education, heuristic solutions and toolbars and plugins. The thesis shows the flaws and improvements in the different solutions. The conclusion in this thesis is that there is no simple answer but that the combination of education and technical solution may be the best one. Technical protection itself can reduce the risk of getting a phishing email, but if the email arrives, there must be awareness and knowledge in how to see the difference between a legitimate email and a phishing email.

  • Meldrum, Max
    KTH, School of Electrical Engineering and Computer Science (EECS).
    Hardware Utilisation Techniques for Data Stream Processing2019Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
    Abstract [en]

    Recent years have seen an increase in use of the stream processing architecture to compose continuous analytics applications. This thesis presents the design of a Rust-based stream processor that adopts two separate techniques to tackle existing weaknesses in modern production-grade stream processors. The first technique employs a data analytics language on top of the streaming runtime, in order to provide both dataflow and low-level compiler optimisations. This technique is motivated by an analysis of the impact that the lack of compiler integration may have on the end-to-end performance of streaming pipelines in Apache Flink. In the second technique streaming operators are scheduled using a task-parallel approach to boost performance for skewed data distributions. The experimental results for data-parallel streaming pipelines in this thesis demonstrate, that the scheduling model of the prototype achieves performance improvements in skewed scenarios without exhibiting any significant losses in performance during uniform distributions.