Bakalauriniame darbe konstruojamas gaminys - „Audinių skersinio pjaustymo įrenginys". Nagrinėti šioje srityje naudojami įtaisai ir priemonės, parinkta keletas galimų konstrukcinių variantų, atliktas jų palyginimas (privalumai ir trūkumai). Atlikti šio prietaiso, transporterio techniniai skaičiavimai. Analizuoti du prietaiso detalių pagaminimo technologiniai maršrutai, nustatyti įrenginiai reikalingi atlikti mechanines operacijas. Darbe pateikiami techniniai ir eksploataciniai reikalavimai naudojant įrenginį darbo procese. Ekonominiais skaičiavimais nustatyta ir pagrįsta gaminio savikaina. / All cutting machinery is similar to each other because of their working principle. They all have cutting knife and some kind of shafts which pulls fabric. „Audinių skersinio pjaustymo įrenginys“ is made from laser cut metal and some construction tubes. Other used drives is electric motors, reducers, pneumatic cylinders, belt drives. This device can make all kind of blankets varying from 10 to 60mm thickness, all sort of fabrics, up to 2000 mm width and custom lenght. This paper describes the purpose of a blanket cutting machine, presents its technical specification and carried out analysis of constructional variants, technical calculations, and operational requirements. When designing the blanket cutting machine information from technical library of joint stock company "UAB Dominari" and collected information on peculiarities of maintenance and repairs of technological equipment of the company were used. Additionally, material available from Šiauliai University Technology Faculty library and methodological material from the department were used, scholarly and technical literature was analyzed, patent search was done, internet search engine was used.

This paper presents the scan-line algorithm which has been implemented on the Lisp Machine. The scan-line algorithm resides beneath a library of primitive software routines which draw more fundamental objects: lines, triangles and rectangles. This routine, implemented in microcode, applies the A(BC)*D approach to word boundary alignments in order to create an extremely fast, efficient, and general purpose drawing primitive. The scan-line algorithm improves on previous methodologies by limiting the number of CPU intensive instructions and by minimizing the number of words referenced. This paper will describe how to draw scan-lines and the constraints imposed upon the scan-line algorithm by the Lisp Machine's hardware and software.

This master thesis is dealing with the design of the frame and working axis of the desktop milling machine. The theoretical part is in the form of a research of the construction of small machine tools. Practical part concerns the idea for a design of the working axis, as well as the design of the frame construction. Practical part also contains the selection of the main and most important components.

The subject of this master’s thesis is a design of a single-purpose assembly machine for automotive industry. The assembled component is a part of a headlight of a passenger car consisting of a heatsink, a PCB and a reflector which are mutually connected by screws. The result of the thesis is detailed 3D model of the single-purpose machine made in Creo Parametric software and drawing documentation of the designed machine.

The diploma thesis contains an overview of the development of numerical control of machine tools and a historical excursion of the beginnings of numerical control. The work deals with the design and presentation of the component, and then the calculation documentation used to determine the load and safety of the component. After computational and design analysis, the work deals with the design of component production on CNC machines and the technological process of production. Finally, the production of the part is documented and evaluated.

Machine translation (MT) as a field of research has known significant advances in recent years, with the increased interest for neural machine translation (NMT). By combining deep learning with translation, researchers have been able to deliver systems that perform better than most, if not all, of their predecessors. While the general consensus regarding NMT is that it renders higher-quality translations that are overall more idiomatic, researchers recognize that NMT systems still struggle to deal with certain classic difficulties, and that their performance may vary depending on their architecture. In this project, we implement a challenge-set based approach to the evaluation of examples of three main NMT architectures: convolutional neural network-based systems (CNN), recurrent neural network-based (RNN) systems, and attention-based systems, trained on the same data set for English to French translation. The challenge set focuses on a selection of lexical and syntactic difficulties (e.g., ambiguities) drawn from literature on human translation, machine translation, and writing for translation, and also includes variations in sentence lengths and structures that are recognized as sources of difficulties even for NMT systems. This set allows us to evaluate performance in multiple areas of difficulty for the systems overall, as well as to evaluate any differences between architectures’ performance. Through our challenge set, we found that our CNN-based system tends to reword sentences, sometimes shifting their meaning, while our RNN-based system seems to perform better when provided with a larger context, and our attention-based system seems to struggle the longer a sentence becomes.

While machine learning has shown success in many fields, it can be challenging when there are limitations with insufficient training data. By incorporating knowledge into the machine learning pipeline, one can overcome such limitations. Therefore, eliciting expert knowledge can play an important role in the machine learning project pipeline. Expert knowledge can come in many forms, and it is seldom easy to elicit and formalize it in a way that is easily implementable into a machine learning project. While it has been done, not much focus has been on how. Furthermore, the motivations for why knowledge was elicited in a particular way as well as the challenges that may exist with the elicitation, are not always focused on either. Making educated decisions for knowledge elicitation can therefore be challenging for researchers. Hence, this work aims to explore and categorize how expert knowledge elicitation has been done by researchers previously. This was done by developing a taxonomy that was then used for analyzing articles. A total of 43 articles were found, containing 97 elicitation paths that were categorized in order to identify trends and common approaches. The findings from our study were used to provide guidance for an industrial case in its initial stage to show how the taxonomy presented in this work can be applied in a real-world scenario.

The demand for ultra-high-speed machines (UHSM) is rapidly growing in high-tech industries due to their attractive features. A-mechanically-based-antenna (AMEBA) system is another emerging application of UHSM. It enables portable wireless communication in the radio frequency (RF)-denied environment, which was not possible until recently. The AMEBA system requires a high-power (HP) UHSM for its effective communication performance. However, at the expected rotational speed range of 0.5 to 1 million rpm, the power level of UHSM is limited, and no research effort has succeeded to improve the power level of UHSM. The design of HP-UHSM is highly iterative, and it presents several critical challenges, unlike low-power UHSM, such as critical-bending-resonance (CBR), strong mutual influence among Multiphysics performances, exponential air-friction loss, and material limitation. When the magnetic loading of the UHSM rotor is increased to improve the power level, the rotor experiences serious mechanical vibration due to the excessive centrifugal forces and CBR. This vibration limits the operation of HP-UHSM and leads to structural breakdown. Furthermore, the design process becomes more critical when it considers the multidisciplinary design constraints and application requirements. This dissertation proposed a new Multiphysics design method to develop HP-UHSM for critical applications. First, the critical design constraints which prevent increasing the output power of UHSM are investigated. Then, a Multiphysics optimization model is developed by coupling several multidisciplinary analysis modules. This proposed optimization model enables (i) defining multidisciplinary design constraints, (ii) consideration of Multiphysics mutual influence, and (iii) a trade-off analysis between the efficiency and design-safety-margin. The proposed design model adopts the multiphase winding system to effectively increase the electrical loading in the slotless stator. Finally, a 2000 W 500,000 rpm HP-UHSM is optimized for an AMEBA system using the proposed design method. The optimized 2 kW 500,000 rpm machine prototype and its dynamo setup are built in the laboratory. Extensive finite element simulations and experimental testing results are presented to validate the effectiveness of the proposed design method. The results show that the proposed HP-USHM has 94.5% efficiency, 47 kW/L power density, 30% global design safety margin at the maximum speed and no CBR frequency below 11 kHz.

[pt] Este trabalho considera o problema de Regressão com Tempo Limitado. Dados um dataset, um algoritmo de aprendizado (Learner) a ser treinado e um tempo limitado, não sabemos se seria possível treinar o modelo com todo o dataset dentro deste tempo. Queremos então elaborar a estratégia que extraia o melhor modelo possível deste algoritmo de aprendizado respeitando o limite de tempo. Uma estratégia consiste em interagir com o Learner de duas formas: enviando exemplos para o Learner treinar e enviando exemplos para o Learner rotular. Nós definimos o que é o problema de Regressão com Tempo Limitado, decompomos o problema de elaborar uma estratégia em subproblemas mais simples e bem definidos, elaboramos uma estratégia natural baseada em escolha aleatória de exemplos e finalmente apresentamos uma estratégia, TW+BH, que supera a estratégia natural em experimentos que realizamos com diversos datasets reais. / [en] This work considers the Time-Limited Regression problem. Given a dataset, a learning algorithm (Learner) to be trained and a limited time, we do not know if it s going to be possible to train the model with the entire dataset within this time constraint. We then want to elaborate the strategy that extracts the best possible model from this learning algorithm respecting the time limit. A strategy consists of a series of interactions with the Learner, in two possible ways: sending labeled examples for the Learner to train and sending unlabeled examples for the Learner to classify. We define what the Time-Limited Regression problem is, we decompose the problem of elaborating a strategy into simpler and more well-defined sub-problems, we elaborate a natural strategy based on random choice of examples and finally we present a strategy, TW+BH, that performs better than the natural strategy in experiments we have done with several real datasets.

With the Internet of Things paradigm, the data generated by the rapidly increasing number of connected devices lead to new possibilities, such as using machine learning for activity recognition in smart environments. However, it also introduces several challenges. The sensors of different devices might be of different types, making the fusion of data non-trivial. Moreover, the devices are often mobile, resulting in that data from a particular sensor is not always available, i.e. there is a need to handle data from a dynamic set of sensors. From a machine learning perspective, the data from the sensors arrives in a streaming fashion, i.e., online learning, as compared to many learning problems where a static dataset is assumed. Machine learning is in many cases a good approach for classification problems, but the performance is often linked to the quality of the data. Having a good data set to train a model can be an issue in general, due to the often costly process of annotating the data. With dynamic and heterogeneous data, annotation can be even more problematic, because of the ever-changing environment. This means that there might not be any, or a very small amount of, annotated data to train the model on at the start of learning, often referred to as the cold start problem. To be able to handle these issues, adaptive systems are needed. With adaptive we mean that the model is not static over time, but is updated if there for instance is a change in the environment. By including human-in-the-loop during the learning process, which we refer to as interactive machine learning, the input from users can be utilized to build the model. The type of input used is typically annotations of the data, i.e. user input in the form of correctly labelled data points. Generally, it is assumed that the user always provides correct labels in accordance with the chosen interactive learning strategy. In many real-world applications these assumptions are not realistic however, as users might provide incorrect labels or not provide labels at all in line with the chosen strategy. In this thesis we explore which interactive learning strategies are possible in the given scenario and how they affect performance, as well as the effect of machine learning algorithms on performance. We also study how a user who is not always reliable, i.e. that does not always provide a correct label when expected to, can affect performance. We propose a taxonomy of interactive online machine learning strategies and test how the different strategies affect performance through experiments on multiple datasets. The findings show that the overall best performing interactive learning strategy is one where the user provides labels when previous estimations have been incorrect, but that the best performing machine learning algorithm depends on the problem scenario. The experiments also show that a decreased reliability of the user leads to decreased performance, especially when there is a limited amount of labelled data.