1 - 4 of 4
rss atomLink to result list
Permanent link
Cite
Citation style
  • apa
  • harvard1
  • 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
  • Presentation: 2019-12-09 10:00 V3, Stockholm
    Varnai, Peter
    KTH, School of Electrical Engineering and Computer Science (EECS), Intelligent systems, Decision and Control Systems (Automatic Control).
    Reinforcement Learning Endowed Robot Planning under Spatiotemporal Logic Specifications2019Licentiate thesis, monograph (Other academic)
    Abstract [en]

    Recent advances in artificial intelligence are producing fascinating results in the field of computer science. Motivated by these successes, the desire to transfer and implement learning methods on real-life systems is growing as well. The increased level of autonomy and intelligence of the resulting systems in carrying out complex tasks can be expected to revolutionize both the industry and our everyday lives. This thesis takes a step towards this goal by studying reinforcement learning methods for solving optimal control problems with task satisfaction constraints. More specifically, spatiotemporal tasks given in the expressive language of signal temporal logic are considered.

    We begin by introducing our proposed solution to the task constrained optimal control problem, which is based on blending traditional control methods with more recent, data-driven approaches. We adopt the paradigm that the two approaches should be considered as endpoints of a continuous spectrum, and incorporate partial knowledge of system dynamics into the learning process in the form of guidance controllers. These guidance controllers aid in enforcing the task satisfaction constraint, allowing the system to explore towards finding optimal trajectories in a more sample-efficient manner. The proposed solution algorithm is termed guided policy improvement with path integrals (G-PI2). We also propose a framework for deriving effective guidance controllers, and the importance of this guidance is illustrated through a simulation case study.

    The thesis also considers a diverse range of enhancements to the developed G-PI2 algorithm. First, the effectiveness of the guidance laws is increased by continuously updating their parameters throughout the learning process using so-called funnel adaptation. Second, we explore a learning framework for gathering and storing experiences gained from previously solved problems in order to efficiently tackle changes in initial conditions or task specifications in future missions. Finally, we look at how so-called robustness metrics, which quantify the extent of task satisfaction for signal temporal logic, can be explicitly defined in order to aid the learning process towards finding task satisfying trajectories. The multidisciplinary nature of the examined task constrained optimal control problem offers a broad range of additional research directions to consider in future work, which are discussed in detail as well.

  • Presentation: 2019-12-10 10:00 M311, Stockholm
    Theissen, Nikolas Alexander
    KTH, School of Industrial Engineering and Management (ITM).
    Physics-based modelling and measurement of advanced manufacturing machinery’s positioning accuracy: Machine tools, industrial manipulators and their positioning accuracy2019Licentiate thesis, comprehensive summary (Other academic)
    Abstract [en]

    Advanced manufacturing machinery is a corner stone of essential industries of technologicallydeveloped societies. Their accuracy permits the production of complexproducts according to tight geometric dimensions and tolerances for high efficiency,interchangeability and sustainability. The accuracy of advanced manufacturingmachinery can be quantified by the performance measure of positioning accuracy.Positioning accuracy measures the closeness between a commanded and an attainedposition on a machine tool or industrial manipulator, and it is ruled by lawsof physics in classical mechanics and thermodynamics. These laws can be applied tomodel how much the machinery deflects due to gravity, expands due to a change intemperature and how much and how long it vibrates due to process forces; hence,one can quantify how much the accuracy decreases. Thus, to produce machinerywith ever higher accuracy and precision one can design machines which deflect,expand and vibrate less or one can understand and model the actual behaviour ofthe machinery to compensate for it.This licentiate thesis uses physics-based modelling to quantify the positioningaccuracy of machine tools and industrial robots. The work investigates the potentialincrease in positioning accuracy because of the simultaneous modelling of the kinematics,static deflections, vibrations and thermo-elasticity as a lumped-parametermodel of the machinery. Consequently the models can be used to quantify thechange of the accuracy throughout the workspace.The lumped parameter models presented in this work require empirical modelcalibration and validation. The success of both, calibration and validation, dependson the availability of the right measurement instruments, as these need to be ableto capture the actual positioning accuracy of machinery. This thesis focuses on theimportance of measurement instruments in industry and metrology and creates acatalogue of requirements and trends to identify the features of the measurementinstruments required for the factories of the future. These novel measurement instrumentsshall be able to improve model calibration and validation for an improvedoverall equipment effectiveness, improved product quality, reduced costs, improvedsafety and sustainability as a result of physics-based modelling and measurementof advanced manufacturing machinery.

  • Presentation: 2019-12-10 14:00 FB54, AlbaNova universitetscentrum, Stockholm
    Ignas, Mickus
    KTH, School of Engineering Sciences (SCI), Physics, Nuclear Engineering.
    Response Matrix Reloaded: for Monte Carlo Simulations in Reactor Physics2019Licentiate thesis, comprehensive summary (Other academic)
    Abstract [en]

    This thesis investigates Monte Carlo methods applied to criticality and time-dependent problems in reactor physics. Due to their accuracy and flexibility, Monte Carlo methods are considered as a “gold standard” in reactor physics calculations. However, the benefits come at a significant computing cost. Despite the continuous rise in easily accessible computing power, a brute-force Monte Carlo calculation of some problems is still beyond the reach of routine reactor physics analyses. The two papers on which this thesis is based try to address the computing cost issue, by proposing methods for performing Monte Carlo reactor physics calculations more efficiently. The first method addresses the efficiency of the widely-used k-eigenvalue Monte Carlo criticality calculations. It suggests, that the calculation efficiency can be increased through a gradual increase of the neutron population size simulated during each criticality cycle, and proposes a way to determine the optimal neutron population size. The second method addresses the application of Monte Carlo calculations to reactor transient problems. While reactor transient calculations can, in principle, be performed using only Monte Carlo methods, such calculations take multiple thousands of CPU hours for calculating several seconds of a transient. The proposed method offers a middle-ground approach, using a hybrid stochastic-deterministic scheme based on the response matrix formalism. Previously, the response matrix formalism was mainly considered for steady-state problems, with limited application to time-dependent problems. This thesis proposes a novel way of using information from Monte Carlo criticality calculations for solving time-dependent problems via the response matrix.

     

  • Presentation: 2019-12-18 10:00 Sal Sefström, Stockholm
    Jarnerud, Tova
    KTH, School of Industrial Engineering and Management (ITM), Materials Science and Engineering.
    Utilization of recovered lime-containing materials from pulp and paper industries as slag formers in stainless steel production2019Licentiate thesis, comprehensive summary (Other academic)
    Abstract [en]

    In recent years, major efforts have been made to develop a more circular economy. The desire to reuse, remanufacture and recycle materials are important for the development towards a sustainable society. An extended lifespan of materials helps to reduce the amount of waste kept in landfills, as well as to reduce the extraction of natural resources. Pulp and paper and steel industries are two of the largest export industrial sectors in Sweden. It is well known that the pulp and paper industries also generates large amounts of organic and inorganic wastes, of which a significant part is kept in landfills. Year by year deposit of wastes in landfills is becoming more difficult to handle and expensive due to stronger regulations and requirements regarding the environment. During Electric Arc Furnace (EAF) and Argon Oxygen Decarburization (AOD) stainless steelmaking operations, burnt lime (primary lime) is charged together with other slag forming materials in the furnace or converter to attain a specific basicity of the slag and to achieve purification from unwanted elements by chemical reactions in the steel.

    However, a number of CaO-containing wastes from pulp/paper mills can be used as slag formers in steelmaking processes to replace the currently used burnt natural lime, since the use of this primary lime does not conform with the closing the loop idea that is being prioritized in modern society. This thesis presents results from preliminary experiments for examining the replacement of primary lime with secondary lime from pulp and paper production waste as slag formers in EAF and AOD converters. The obtained results showed a possibility of using up to 30% CaO from secondary lime as a replacement for primary lime in the EAF. Furthermore, the amount of ferrosilicon alloys additions can be decreased by up to 3kg/ton of steel. For the AOD process the use of secondary lime doesn´t have any negative effects on the decarburization process and reduction process. Furthermore, it has similar desulphurization functions as the primary lime. However, the phosphorus content in the metal was slightly increased. Thus, the replacement ratio of secondary lime will be limited by the acceptable phosphorus level in the final steel due to higher phosphorus content in wastes from pulp and paper mills compared to that in primary lime. Moreover, it was revealed during this study that slags from AOD converters can be used as binding agents for briquetting of these secondary lime materials. These results shows that waste/by products from two major industries can be used to make metallurgical briquettes, uniting two major industrial sectors in a circular symbiosis towards a more sustainable future.