| 1. |
- A Asif, Farazee M, 1980-
(author)
-
Circular Manufacturing Systems : A development framework with analysis methods and tools for implementation
- 2017
-
Doctoral thesis (other academic/artistic)abstract
- The society today lives on the philosophy of ‘take-make-use-dispose.’ In the long run, this is not sustainable as the natural resources and the waste carrying capacity of the earth are limited. Therefore, it is essential to reduce dependency on the natural resources by decoupling the growth from the consumption. In this venture, both the society and the manufacturing industry have a vital role to play. The society needs to shift towards Circular Economy that rests upon the philosophy of ‘take-make-use-reuse’ and the manufacturing industry has to be a major stakeholder in this shift. Despite being proven to be both economically and environmentally beneficial, successful examples of circular systems are few today. This is primarily due to two reasons; firstly, there is a lack of systemic and systematic approach to guide industries and secondly, there is a lack of analysis methods and tools that are capable of assessing different aspects of circular manufacturing systems. Taking on to these challenges, the objective of this research is to bring forward a framework with methods and decision support tools that are essential to implement circular manufacturing systems. The initial conceptual framework with the systemic approach is developed based on extensive review and analysis of research, which is further adapted for industrial implementation. Systematic analysis methods, decision support and implementation tools are developed to facilitate this adaptation. This development has been supported by four cases from diverse manufacturing sectors. Behind each decision support tool, there are analysis methods built upon mainly system dynamics principles. These tools are based on simulation platforms called Stella and Anylogic. Among other things, these tools are capable of assessing the performance of closed-loop supply chains, consequences of resource scarcity, potential gains from resource conservation and overall economic and environmental performance of circular manufacturing systems.
|
|
| 2. |
- A. M. Naiini, Maziar, 1980-
(author)
-
Horizontal Slot Waveguides for Silicon Photonics Back-End Integration
- 2014
-
Doctoral thesis (other academic/artistic)abstract
- This thesis presents the development of integrated silicon photonic devices. These devices are compatible with the present and near future CMOS technology. High-khorizontal grating couplers and waveguides are proposed. This work consists of simulations and device design, as well as the layout for the fabrication process, device fabrication, process development, characterization instrument development and electro-optical characterizations.The work demonstrates an alternative solution to costly silicon-on-insulator photonics. The proposed solution uses bulk silicon wafers and thin film deposited waveguides. Back-end deposited horizontal slot grating couplers and waveguides are realized by multi-layers of amorphous silicon and high-k materials.The achievements of this work include: A theoretical study of fully etched slot grating couplers with Al2O3, HfO2 and AIN, an optical study of the high-k films with spectroscopic ellipsometry, an experimental demonstration of fully etched SiO2 single slot grating couplers and double slot Al2O3 grating couplers, a practical demonstration of horizontal double slot high-k waveguides, partially etched Al2O3 single slot grating couplers, a study of a scheme for integration of the double slot Al2O3 waveguides with selectively grown germanium PIN photodetectors, realization of test chips for the integrated germanium photodetectors, and study of integration with graphene photodetectors through embedding the graphene into a high-k slot layer.From an application point of view, these high-k slot waveguides add more functionality to the current silicon photonics. The presented devices can be used for low cost photonics applications. Also alternative optical materials can be used in the context of this photonics platform.With the robust design, the grating couplers result in improved yield and a more cost effective solution is realized for integration of the waveguides with the germanium and graphene photodetectors.
|
|
| 3. |
- A Monfared, Behzad, 1983-
(author)
-
Magnetic Refrigeration for Near Room-Temperature Applications
- 2018
-
Doctoral thesis (other academic/artistic)abstract
- Refrigeration plays a crucial role in many different sectors and consumes about 17% of the electricity produced globally. This significant energy consumption implies large share of refrigeration in primary energy consumption and other environmental impacts. In addition to the environmental impacts associated with energy consumption, the vapor-compression systems contribute in global warming due to the release of their gaseous refrigerants into the atmosphere. As an alternative technology for near room-temperature applications, magnetic refrigeration is proposed by some researchers to eliminate the release of gaseous refrigerants into the atmosphere and to reduce the energy consumption. This thesis is a compilation of a number of studies done on magnetic refrigeration for room-temperature applications.In the first study, the environmental impacts associated to magnetic refrigeration are looked at closely through a life cycle assessment. The life cycle assessment indicates that because of the environmental burdens related to the rare-earth materials used in magnetic refrigeration, the reduction in the environmental impacts is not guaranteed by switching to magnetic refrigeration technology. Accordingly to avoid the extra environmental impacts the magnetic refrigeration systems should use magnetic materials frugally, which requires an optimized design. In addition, operation with higher efficiency compared to vapor-compression systems is necessary to have environmental advantages, at least in some impact categories.A practical method to optimize the design of magnetic refrigeration systems, e.g. to have a compact design or high efficiency, is utilizing a flexible software model, with which the effect of varying different parameters on the performance of the system can be simulated. Such a software model of the magnetic refrigeration system is developed and validated in this project. In developing the model one goal is to add to the precision of the simulated results by taking more details into consideration. This goal is achieved by an innovative way of modeling the parasitic heat transfer and including the effect of the presence of magnetocaloric materials on the strength of the field created by the magnet assembly. In addition, some efforts are made to modify or correct the existing correlations to include the effect of binding agents used in some active magnetic regenerators. Validation of the developed software model is done using the experimental results obtained from the prototype existing at the Department of Energy Technology, KTH Royal Institute of Technology.One of the parameters that can be modified by the developed software model is the choice of the magnetocaloric materials for each layer in a layered active magnetic regenerator. Utilizing the software model for optimizing the choice of the materials for the layers reveals that materials with critical temperatures equal to the cyclic average temperature of the layers in which they are used do not necessarily result in the desired optimum performance. In addition, for maximizing different outputs of the models, such as energy efficiency or temperature lift sustained at the two ends of the regenerators, different choice of materials for the layers are needed. Therefore, in other studies seeking to improve one of the outputs of a system, the choice of the transition or critical temperatures of the materials for each layer is an additional parameter to be optimized.The prototype existing at the Department of Energy Technology, KTH Royal Institute of Technology, was initially designed for replacing the vapor-compression system of a professional refrigerator. However, it could not fulfil the requirements for which it was initially designed. The aforementioned developed simulation model is used to see how much the choice of the materials, size of the particles, and number of layers can enhance the performance while the operation frequency and flow rate of the heat transfer fluid are at their optimum values. In other words, in that study the room for improvement in the performance without applying major changes in the system such as the geometry of the regenerator, which implies redesigning the whole magnet assembly, is investigated. In the redesign process the effect of binding agent and the limitations associated to different properties of it is also investigated theoretically. Nevertheless, the study did not show that with keeping the geometry of the regenerators and the currently existing magnetocaloric materials the initial goals of the prototype can be achieved.In the next study more flexible choice of geometries and magnetocaloric materials are considered. In fact, in this study it is investigated how much the magnetocaloric materials need to be improved so that magnetic refrigeration systems can compete with vapor-compression ones in terms of performance. For the two investigated cases, the magnetic-field dependent properties of the currently existing materials are enough provided that some other issues such as low mechanical stability and inhomogeneity of the properties are solved. Nevertheless, for more demanding design criteria, such as delivering large cooling capacity over a considerable temperature span while the magnetic materials are used sparingly, the magnetic-field dependent properties need to be enhanced, as well.A less explored area in room-temperature magnetic refrigeration is the subject of another study included in the thesis. In this study, solid-state magnetic refrigeration systems with Peltier elements as heat switches are modeled. Since the Peltier elements consume electricity to pump heat, the modeled systems can be considered hybrid magnetocaloric-Peltier cooling systems. For such systems the detailed transient behavior of the Peltier elements together with layers of magnetocaloric materials are modeled. The mathematical model is suitable for implementation in programing languages without the need for commercial modeling platforms. The parameters affecting the performance of the modeled system are numerous, and optimization of them requires a separate study. However, the preliminary attempts on optimizing the modeled system does not give promising results. Accordingly, focusing on passive heat switches can be more beneficial.
|
|
| 4. |
- Aaltonen, Jaana, 1969-
(author)
-
Ground Monitoring using Resistivity Measurements in Glaciated Terrains
- 2001
-
Doctoral thesis (other academic/artistic)abstract
- The most common method of monitoring and mapping groundwatercontaminants is to extract and analyse a number of groundwatersamples from wells in the investigation area. However, thereare a number of limitations with this type of point-wiseinvestigation, as it is hard to acquire an adequate picture ofa heterogeneous and anisotropic subsurface using a fewpoints. To overcome the limitations of point investigations and toimprove ground monitoring investigations in a cost-effectiveway, support can be provided by direct current resistivitymeasurements, which give a characterisation of the electricalproperties of a ground volume. The main objective with this work was to investigate theusability of the resistivity method as a support in monitoringgroundwater contaminants in glaciated terrains and underdifferent seasons, both in long-term monitoring programmes andintracer tests. The work comprised field investigations at several differentsanitary landfills and four tracer tests in differentgeological environments, around the Stockholm region. The maininvestigations have been done at Högbytorp, Stockholmwhich has been used for long-term investigations of theresistivity variation, together with a field set up formonitoring and measurements on seasonal variation in soilmoisture, ground temperature and precipitation. It can be concluded that the use of resistivity measurementssupplies valuable information in the case of mapping andmonitoring conductive groundwater contaminants andfurthermore: The variation in resistivity (in shallow investigations<1 m) can be extensive between different seasons (around30 % compared to a mean value in till and clay soils) andshould be considered, so that anthropogenic affects can beseparated from natural resistivity variation. For deeperinvestigations (>5 m) the seasonal resistivity variationwas more moderate (around 15% compared to a mean value intill and clay soils). Soil moisture variation shows a strong relationship toresistivity variation in the investigated clay and tillsoils. Together with temperature correction 47 to 65% of thevariation has been explained. Three types of monitoring systems can be applied:Permanently installed, partly installed and fully mobilesystems. For the actual measurements, all three types can useeither high-density techniques such as CVES (ContinuousVertical Electrical Sounding) or low-density measuring withone or some different electrode spacings. The suggested evaluation tool for monitoring programmesshowed that it was possible to detect a decrease of 15 % inthe mean value at a specific site using Modified Double Masscalculations between resistivity time series and time seriesat a reference site with a comparable seasonalvariation. Resistivity measurements may be used as a valuablecomplement to groundwater sampling in tracer tests. Adecrease in resistivity, a minimum and a recovery phasereflect the passage of a NaCl-solution, which can be used toestimate flow velocity and flow patterns of the investigatedaquifer. The achieved recovery of NaCl in the tracer testscarried out was estimated to 20 to 70 %. The measurement system for long-term monitoring or tracertests, which should be chosen with regard to layout andfrequency, depends on the purpose of measurement and onsite-specific conditions and therefore no standard solutioncan be proposed. Key words:Resistivity, Direct Current, Monitoring,Groundwater, Contaminant, Tracer test, Geophysics.
|
|
| 5. |
- Aare, Magnus, 1974-
(author)
-
Prevention of Head Injuries - focusing Specifically on Oblique Impacts
- 2003
-
Doctoral thesis (other academic/artistic)abstract
- The massive number of injuries sustained in trafficaccidents is a growing problem worldwide, especially indeveloping countries. In 1998, more than one million peoplewere killed in traffic accidents worldwide, while about tentimes as many people were injured. Injuries to the centralnervous system and in particular to the headare especiallycritical to human life. This thesis contains five researchpapers looking at head injuries and head protection, proposingnew and more efficient ways of protecting the head, especiallyin traffic accidents.In order to define the national dimensions of the patternsof injuries incurred in motorcycle and moped accidents inSweden, a statistical survey was performed on data spanning a13-year period (Paper A). In Sweden, 27,100 individualsreceived in-patient care for motorcycle and moped accidentinjuries between 1987 and 1999. The motorcycle and moped injuryrate reduced in the second half of the study period, so toowere the total number of days of treatment per year. Males hadeight times the incidence of injuries of females. Head injurieswere the single most frequent diagnosis, followed by fracturesof the lower limbs. Concussion was the most frequent headinjury. These statistics clearly show the need for better headinjury prevention systems.According to the statistics, the most common type of impactto the head in motorcycle and moped accidents is an obliqueimpact. Oblique impacts generate rotations of the head, whichare a common cause of the most severe head injuries. Thereforea new test rig was constructed to reproduce oblique impacts toa helmeted dummy head, simulating those occurring in real lifeaccidents (Paper B). The new test rig was shown to provideuseful data at speeds of up to 50 km/h and with impact anglesvarying from purely tangential to purely radial. Thisinnovative test rig appears to provide an accurate method formeasuring accelerations in oblique impacts to helmets.When testing the performances of motorcycle helmets,discrepancies are usually seen in the test results. In order toevaluate these discrepancies, the finite element method (FEM)was used for simulations of a few oblique helmet impacts (PaperC). Amongthe parameters studied, the coefficients of frictionbetween the impacting surface and the helmet and between thehead and the helmet had the most significant influence on therotational accelerations. Additionally, a thinner andconsequently also weaker shell and a weaker liner, providedbetter protection for the impacts studied.Since there are no generally accepted global injurythresholds for oblique impacts to the human head, a study wasdesigned to propose new injury tolerances accounting for bothtranslations and rotations of the head (Paper D). In thatstudy, FE models of (a) a human head, (b) a Hybrid III dummyhead, and (c) the experimental helmet were used. Differentcriteria were proposed for different impact scenarios. Both thetranslational and the rotational effects were found to beimportant when proposing a predictor equation for the strainlevels experienced by the human brain in simulated impacts tothe head.In order to reduce the level of head injuries in society andto better understand helmet impacts from different aspect, aballistic impact was also studied (Paper E). The effects ofdifferent helmet shell stiffness and different angles ofimpacts were simulated. In this study, the same FE head modelfrom Paper D was used, however here it was protected with amodel of a composite ballistic helmet. It was concluded thatthe helmet shell should be stiff enough to prevent the insideof the shell from striking the skull, and that the strainsarising in the brain tissue were higher for some obliqueimpacts than for purely radial ones.In conclusion, this thesis describes the injury pattern ofmotorcycle and moped accidents in Sweden. This thesis showsthat the injuries sustained from these accidents can bereduced. In order to study both translational as well asrotational impacts, a new laboratory test rig was designed. Byusing the finite element method, it is possible to simulaterealistic impacts to the head and also to predict how severehead injuries may potentially be prevented.
|
|
| 6. |
- Aas, Erik, 1990-
(author)
-
A Markov Process on Cyclic Words
- 2014
-
Doctoral thesis (other academic/artistic)abstract
- The TASEP (totally asymmetric simple exclusion process) studied here is a Markov chain on cyclic words over the alphabet{1,2,...,n} given by at each time step sorting an adjacent pair of letters chosen uniformly at random. For example, from the word 3124 one may go to 1324, 3124, 3124, 4123 by sorting the pair 31, 12, 24, or 43.Two words have the sametype if they are permutations of each other. If we restrict TASEP to words of some particular type m we get an ergodic Markov chain whose stationary distribution we denote by ζm. Soζm (u) is the asymptotic proportion of time spent in the state u if the chain started in some word of type m. The distribution ζ is the main object of study in this thesis. This distribution turns out to have several remarkable properties, and alternative characterizations. It has previously been studied both from physical, combinatorial, and probabilitistic viewpoints.In the first chapter we give an extended summary of known results and results in this thesis concerning ζ. The new results are described (and proved) in detail in Papers I - IV.The new results in Papers I and II include an explicit formula for the value ofζat sorted words and a product formula for decomposable words. We also compute some correlation functions for ζ. In Paper III we study of a generalization of TASEP to Weyl groups. In Paper IV we study a certain scaling limit of ζ, finding several interesting patterns of which we prove some. We also study an inhomogenous version of TASEP, in which different particles get sorted at different rates, which generalizes the homogenous version in several aspects. In the first chapter we compute some correlation functions for ζ
|
|
| 7. |
- Aas, Wenche
(author)
-
Constitution, dynamics and structure of binary and ternary actinide complexes
- 1999
-
Doctoral thesis (other academic/artistic)abstract
- Stoichiometry, ligand exchange reactions, coordinationgeometry and stability of complexes of type UO2LpFq(H2O)3-n(p= 1–2,q= 1-3), where L is one of the bidentate ligandspicolinate, oxalate, carbonate or acetate have beeninvestigated using single crystal X-ray diffraction, an arrayof19F-,13C-,17O- and1H-NMR techniques and potentiometric titration usingboth F-and H+selective electrodes. The experiments wereperformed in a 1.00 M NaClO4medium. The equilibrium constants were determinedat 25°C while most of the kinetic experiments were done at- 5°C. The equilibrium constants for the stepwise additionof F-to UO2L and UO2L2indicates that the prior coordination of L toU(VI) has a fairly small effect on the subsequent bonding offluoride, except for a statistical effect determined by thenumber of available coordination sites. This indicates thatternary complexes might be important for the speciation andtransport of hexavalent actinides in ground and surface watersystems. A single crystal structure of UO2(picolinate)F32-has been determined showing the same pentagonalbipyramidal symmetry as in aqueous solution studied by NMR. Theexchangeable donor atoms are situated in a plane perpendicularto the linear uranyl group. The complexes show a variety ofdifferent exchange reactions depending on the ligand used. Ithas been possible to quantify external fluoride and the otherligands exchange reactions as well asintra-molecular reactions. This type of detailedinformation has not been observed in aqueous solution before.Water takes a critical part in the exchange mechanism, and whenit is eliminated from the inner coordination sphere a muchslower kinetics can be observed.19F-NMR has showed to be a powerful technique tostudy these reactions, both because of the sensitivity of thisNMR nucleus and also the possibility to observe reactions wherefluoride is not directly involved in the mechanism. TernaryTh(edta)F1-2and (UO2)2(edta)2F1-4have been investigated using1H and19F-NMR. The fluoride complexation to Cm(III) wasstudied using time resolved fluorescence spectroscopy (TRLFS)and the stability constant for the CmF2+complex was determined at 25°C in 1.0 mNaCl.Keywords. Ternary complexes, actinides,dioxouranium(VI), curium(III), thorium(IV), ligand exchange,isomers, NMR, potentiometric titrations, aqueous solution,oxalate, picolinate, acetate, EDTA.
|
|
| 8. |
- Aastrup, Teodor
(author)
-
In situ investigations of the metal/atmophere interface
- 1999
-
Doctoral thesis (other academic/artistic)abstract
- This thesis presents a new experimental setup for surfaceanalysis studies, incorporating chemical information frominfrared reflection absorption spectroscopy (IRAS) andquantitative information from quartz crystal microbalance(QCM). The integrated IRAS/QCM surface analysis system enablesin situstudies to be conducted of the metal/atmosphereinteraction on a sub-monolayer level. In combination withtopographical information from tapping mode atomic forcemicroscopy (TM-AFM), it was possible to follow the formation ofcorrosion products providing information about chemicalidentity, quantity and lateral distribution. The newlydeveloped system was used to study the influence of relativehumidity (RH) and levels of sulfur dioxide (SO2), ozone (O3)and nitrogen dioxide (NO2) down to concentrations of subparts-per-million (ppm), on the initial stages of atmosphericcorrosion of copper.Exposure to RH led to the formation of small grains ofcuprous oxide (Cu2O), with an average diameter of approximately40 nm. The grain nucleation rate increased with increasing RHdue to the increasing amount of physisorbed water on thesurface.Exposure to RH and SO2 led to the formation of coppersulfite (CuSO3•xH2O) species as well as reduction in therate of Cu2O formation. The formation of CuSO3•xH2Ooccurred through the dissolution and consecutive precipitationof surface-bound copper sulfite complexes.The roles of O3 and NO2 in the SO2-induced atmosphericcorrosion of copper was also investigated. O3 was found to havea two-fold effect, resulting in the oxidation of the coppersulfite species to copper sulfate and the enhancement of Cu2Oformation. NO2 promoted the oxidation of copper sulfitespecies, though less efficiently than O3, and reduced Cu2Oformation. In addition, copper nitrate formation wasobserved.IRAS was also used to investigate the interaction betweenSO2 and surfaces made of platinum and palladium. Sulfuric acidformation was observed on both metals, showing the importanceof the surface on the formation of the metal/atmospherereaction products.Keywords:Atmospheric corrosion, IR-spectroscopy, IRAS,QCM, AFM, SO2, O3, NO2, Cu.
|
|
| 9. |
- Abbas, Haider, 1979-
(author)
-
Options-Based Security-Oriented Framework for Addressing Uncerainty Issues in IT Security
- 2010
-
Doctoral thesis (other academic/artistic)abstract
- Continuous development and innovation in Information Technology introduces novel configuration methods, software development tools and hardware components. This steady state of flux is very desirable as it improves productivity and the overall quality of life in societies. However, the same phenomenon also gives rise to unseen threats, vulnerabilities and security concerns that are becoming more critical with the passage of time. As an implication, technological progress strongly impacts organizations’ existing information security methods, policies and techniques, making obsolete existing security measures and mandating reevaluation, which results in an uncertain IT infrastructure. In order to address these critical concerns, an options-based reasoning borrowed from corporate finance is proposed and adapted for evaluation of security architecture and decision- making to handle them at organizational level. Options theory has provided significant guidance for uncertainty management in several domains, such as Oil & Gas, government R&D and IT security investment projects. We have applied options valuation technique in a different context to formalize optimal solutions in uncertain situations for three specific and identified uncertainty issues in IT security. In the research process, we formulated an adaptation model for expressing options theory in terms useful for IT security which provided knowledge to formulate and propose a framework for addressing uncertainty issues in information security. To validate the efficacy of this proposed framework, we have applied this approach to the SHS (Spridnings- och Hämtningssystem) and ESAM (E-Society) systems used in Sweden. As an ultimate objective of this research, we intend to develop a solution that is amenable to automation for the three main problem areas caused by technological uncertainty in information security: i) dynamically changing security requirements, ii) externalities caused by a security system, iii) obsoleteness of evaluation. The framework is general and capable of dealing with other uncertainty management issues and their solutions, but in this work we primarily deal with the three aforementioned uncertainty problems. The thesis presents an in-depth background and analysis study for a proposed options-based security-oriented framework with case studies for SHS and ESAM systems. It has also been assured that the framework formulation follows the guidelines from industry best practices criteria/metrics. We have also proposed how the whole process can be automated as the next step in development.
|
|
| 10. |
- Abbas, Zainab
(author)
-
Scalable Streaming Graph and Time Series Analysis Using Partitioning and Machine Learning
- 2021
-
Doctoral thesis (other academic/artistic)abstract
- Recent years have witnessed a massive increase in the amount of data generated by the Internet of Things (IoT) and social media. Processing huge amounts of this data poses non-trivial challenges in terms of the hardware and performance requirements of modern-day applications. The data we are dealing with today is of massive scale, high intensity and comes in various forms. MapReduce was a popular and clever choice of handling big data using a distributed programming model, which made the processing of huge volumes of data possible using clusters of commodity machines. However, MapReduce was not a good fit for performing complex tasks, such as graph processing, iterative programs and machine learning. Modern data processing frameworks, that are being popularly used to process complex data and perform complex analysis tasks, overcome the shortcomings of MapReduce. Some of these popular frameworks include Apache Spark for batch and stream processing, Apache Flink for stream processing and Tensor Flow for machine learning.In this thesis, we deal with complex analytics on data modeled as time series, graphs and streams. Time series are commonly used to represent temporal data generated by IoT sensors. Analysing and forecasting time series, i.e. extracting useful characteristics and statistics of data and predicting data, is useful for many fields that include, neuro-physiology, economics, environmental studies, transportation, etc. Another useful data representation we work with, are graphs. Graphs are complex data structures used to represent relational data in the form of vertices and edges. Graphs are present in various application domains, such as recommendation systems, road traffic analytics, web analysis, social media analysis. Due to the increasing size of graph data, a single machine is often not sufficient to process the complete graph. Therefore, the computation, as well as the data, must be distributed. Graph partitioning, the process of dividing graphs into subgraphs, is an essential step in distributed graph processing of large scale graphs because it enables parallel and distributed processing.The majority of data generated from IoT and social media originates as a continuous stream, such as series of events from a social media network, time series generated from sensors, financial transactions, etc. The stream processing paradigm refers to the processing of data streaming that is continuous and possibly unbounded. Combining both graphs and streams leads to an interesting and rather challenging domain of streaming graph analytics. Graph streams refer to data that is modelled as a stream of edges or vertices with adjacency lists representing relations between entities of continuously evolving data generated by a single or multiple data sources. Streaming graph analytics is an emerging research field with great potential due to its capabilities of processing large graph streams with limited amounts of memory and low latency. In this dissertation, we present graph partitioning techniques for scalable streaming graph and time series analysis. First, we present and evaluate the use of data partitioning to enable data parallelism in order to address the challenge of scale in large spatial time series forecasting. We propose a graph partitioning technique for large scale spatial time series forecasting of road traffic as a use-case. Our experimental results on traffic density prediction for real-world sensor dataset using Long Short-Term Memory Neural Networks show that the partitioning-based models take 12x lower training time when run in parallel compared to the unpartitioned model of the entire road infrastructure. Furthermore, the partitioning-based models have 2x lower prediction error (RMSE) compared to the entire road model. Second, we showcase the practical usefulness of streaming graph analytics for large spatial time series analysis with the real-world task of traffic jam detection and reduction. We propose to apply streaming graph analytics by performing useful analytics on traffic data stream at scale with high throughput and low latency. Third, we study, evaluate, and compare the existing state-of-the-art streaming graph partitioning algorithms. We propose a uniform analysis framework built using Apache Flink to evaluate and compare partitioning features and characteristics of streaming graph partitioning methods. Finally, we present GCNSplit, a novel ML-driven streaming graph partitioning solution, that uses a small and constant in-memory state (bounded state) to partition (possibly unbounded) graph streams. Our results demonstrate that \ours provides high-throughput partitioning and can leverage data parallelism to sustain input rates of 100K edges/s. GCNSplit exhibits a partitioning quality, in terms of graph cuts and load balance, that matches that of the state-of-the-art HDRF (High Degree Replicated First) algorithm while storing three orders of magnitude smaller partitioning state.
|
|
