| 1. |
- Calikus, Ece, 1990-
(författare)
-
Self-Monitoring using Joint Human-Machine Learning : Algorithms and Applications
- 2020
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The ability to diagnose deviations and predict faults effectively is an important task in various industrial domains for minimizing costs and productivity loss and also conserving environmental resources. However, the majority of the efforts for diagnostics are still carried out by human experts in a time-consuming and expensive manner. Automated data-driven solutions are needed for continuous monitoring of complex systems over time. On the other hand, domain expertise plays a significant role in developing, evaluating, and improving diagnostics and monitoring functions. Therefore, automatically derived solutions must be able to interact with domain experts by taking advantage of available a priori knowledge and by incorporating their feedback into the learning process.This thesis and appended papers tackle the problem of generating a real-world self-monitoring system for continuous monitoring of machines and operations by developing algorithms that can learn data streams and their relations over time and detect anomalies using joint-human machine learning. Throughout this thesis, we have described a number of different approaches, each designed for the needs of a self-monitoring system, and have composed these methods into a coherent framework. More specifically, we presented a two-layer meta-framework, in which the first layer was concerned with learning appropriate data representations and detectinganomalies in an unsupervised fashion, and the second layer aimed at interactively exploiting available expert knowledge in a joint human-machine learning fashion.Furthermore, district heating has been the focus of this thesis as the application domain with the goal of automatically detecting faults and anomalies by comparing heat demands among different groups of customers. We applied and enriched different methods on this domain, which then contributed to the development and improvement of the meta-framework. The contributions that result from the studies included in this work can be summarized into four categories: (1) exploring different data representations that are suitable for the self-monitoring task based on data characteristics and domain knowledge, (2) discovering patterns and groups in data that describe normal behavior of the monitored system/systems, (3) implementing methods to successfully discriminate anomalies from the normal behavior, and (4) incorporating domain knowledge and expert feedback into self-monitoring.
|
|
| 2. |
- Irgang dos Santos, Luís Fernando, 1989-
(författare)
-
Continuous Finding Problems and Implementing Solutions in Health Care-Associated Infections : The Role of Infection Preventionists
- 2020
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This licentiate thesis aims to understand how infection preventionists (IPs) continuously find problems and implement solutions related to health care-associated infections (HAIs) in hospital settings.HAIs are infections acquired by patients during the process of care and are among the main causes of deaths worldwide. Recently, practices for HAIs prevention and control have challenged IPs due to pandemics (e.g. COVID-19), antimicrobial resistance, population aging and limited resources in health care facilities. Such challenges demand actions to find, solve problems and implement solutions. However, IPs often fail to address these problems. The reasons stem from their inability to timely identify valuable problems and implement new solutions. Although the literature on infection prevention and control is well developed, previous studies have largely investigated how IPs implement preconceived practices to solve given problems as a single event, rather than on how to continuously find problems and implement solutions. This licentiate thesis comprises two empirical papers. Paper I investigates how infection prevention and control teams find problems with HAIs, and is based on a multiple case study of three infection prevention and control teams from one Swedish and two Brazilian hospitals. Paper II investigates how IPs continuously implement changes in infection prevention and control practices during pandemics, and is based on a qualitative descriptive study. The data in both papers were collected from 44 semi-structured interviews with health care professionals enrolled as IPs in Brazilian and Swedish hospitals. The key theories and literatures covered include Problem-Finding and Problem-Solving Perspective and Implementation research.This licentiate thesis contains three main contributions. First, it advances the Problem-Finding and Problem-Solving Perspective literature by providing empirical evidence on how to create valuable knowledge from ill-structured and complex problems. Second, this licentiate thesis suggests a distinction between HAI prevention and HAI control based on two modes of decision-making for finding valuable problems with HAIs. Third, the licentiate thesis describes and categorizes sets of practices that allow to continuously implement changes of infection prevention and control practices during pandemics.
|
|
| 3. |
- KRISHNA, AMOGH VEDANTHA, 1990
(författare)
-
Towards Topography Characterization of Additive Manufacturing Surfaces
- 2020
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Additive Manufacturing (AM) is on the verge of causing a downfall to conventional manufacturing with its huge potential in part manufacture. With an increase in demand for customized product, on-demand production and sustainable manufacturing, AM is gaining a great deal of attention from different industries in recent years. AM is redefining product design by revolutionizing how products are made. AM is extensively utilized in automotive, aerospace, medical and dental applications for its ability to produce intricate and lightweight structures. Despite their popularity, AM has not fully replaced traditional methods with one of the many reasons being inferior surface quality. Surface texture plays a crucial role in the functionality of a component and can cause serious problems to the manufactured parts if left untreated. Therefore, it is necessary to fully understand the surface behavior concerning the factors affecting it to establish control over the surface quality. The challenge with AM is that it generates surfaces that are different compared to conventional manufacturing techniques and varies with respect to different materials, geometries and process parameters. Therefore, AM surfaces often require novel characterization approaches to fully explain the manufacturing process. Most of the previously published work has been broadly based on two-dimensional parametric measurements. Some researchers have already addressed the AM surfaces with areal surface texture parameters but mostly used average parameters for characterization which is still distant from a full surface and functional interpretation. There has been a continual effort in improving the characterization of AM surfaces using different methods and one such effort is presented in this thesis. The primary focus of this thesis is to get a better understanding of AM surfaces to facilitate process control and optimization. For this purpose, the surface texture of Fused Deposition Modeling (FDM) and Laser-based Powder Bed Fusion of Metals (PBF-LB/M) have been characterized using various tools such as Power Spectral Density (PSD), Scale-sensitive fractal analysis based on area-scale relations, feature-based characterization and quantitative characterization by both profile and areal surface texture parameters. A methodology was developed using a Linear multiple regression and a combination of the above-mentioned characterization techniques to identify the most significant parameters for discriminating different surfaces and also to understand the manufacturing process. The results suggest that the developed approaches can be used as a guideline for AM users who are looking to optimize the process for gaining better surface quality and component functionality, as it works effectively in finding the significant parameters representing the unique signatures of the manufacturing process. Future work involves improving the accuracy of the results by implementing improved statistical models and testing other characterization methods to enhance the quality and function of the parts produced by the AM process.
|
|
| 4. |
- Reddy, Vijeth Venkataram, 1990
(författare)
-
On Deterministic feature-based Surface Analysis
- 2020
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Manufacturing sector is continuously identifying opportunities to streamline production, reduce waste and improve manufacturing efficiency without compromising product quality. Continuous improvement has been the primary objective to produce acceptable quality products and meet dynamic customer demands by using advanced techniques and methods. Considering the current demands from society on improving the efficiency with sustainable goals, there is considerable interest from researchers and industry to explore the potential, to optimize- and customize manufactured surfaces, as one way of improving the performance of products and processes. Every manufacturing process generate surfaces which beholds certain signature features. Engineered surfaces consist of both, features that are of interest and features that are irrelevant. These features imparted on the manufactured part vary depending on the process, materials, tooling and manufacturing process variables. Characterization and analysis of deterministic features represented by significant surface parameters helps the understanding of the process and its influence on surface functional properties such as wettability, fluid retention, friction, wear and aesthetic properties such as gloss, matte. In this thesis, a general methodology with a statistical approach is proposed to extract the robust surface parameters that provides deterministic and valuable information on manufactured surfaces. Surface features produced by turning, injection molding and Fused Deposition Modeling (FDM) are characterized by roughness profile parameters and areal surface parameters defined by ISO standards. Multiple regression statistics is used to resolve surfaces produced with multiple process variables and multiple levels. In addition, other statistical methods used to capture the relevant surface parameters for analysis are also discussed in this thesis. The selected significant parameters discriminate between the samples produced by different process variables and helps to identify the influence of each process variable. The discussed statistical approach provides valuable information on the surface function and further helps to interpret the surfaces for process optimization. The research methods used in this study are found to be valid and applicable for different manufacturing processes and can be used to support guidelines for the manufacturing industry focusing on process optimization through surface analysis. With recent advancement in manufacturing technologies such as additive manufacturing, new methodologies like the statistical one used in this thesis is essential to explore new and future possibilities related to surface engineering.
|
|
| 5. |
- Rezk, Nesma, 1987-
(författare)
-
Exploring Efficient Implementations of Deep Learning Applications on Embedded Platforms
- 2020
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- The promising results of deep learning (deep neural network) models in many applications such as speech recognition and computer vision have aroused a need for their realization on embedded platforms. Augmenting DL (Deep Learning) in embedded platforms grants them the support to intelligent tasks in smart homes, mobile phones, and healthcare applications. Deep learning models rely on intensive operations between high precision values. In contrast, embedded platforms have restricted compute and energy budgets. Thus, it is challenging to realize deep learning models on embedded platforms.In this thesis, we define the objectives of implementing deep learning models on embedded platforms. The main objective is to achieve efficient implementations. The implementation should achieve high throughput, preserve low power consumption, and meet real-time requirements.The secondary objective is flexibility. It is not enough to propose an efficient hardware solution for one model. The proposed solution should be flexible to support changes in the model and the application constraints. Thus, the overarching goal of the thesis is to explore flexible methods for efficient realization of deep learning models on embedded platforms.Optimizations are applied to both the DL model and the embedded platform to increase implementation efficiency. To understand the impact of different optimizations, we chose recurrent neural networks (as a class of DL models) and compared its' implementations on embedded platforms. The comparison analyzes the optimizations applied and the corresponding performance to provide conclusions on the most fruitful and essential optimizations. We concluded that it is essential to apply an algorithmic optimization to the model to decrease it's compute and memory requirement, and it is essential to apply a memory-specific optimization to hide the overhead of memory access to achieve high efficiency. Furthermore, it has been revealed that many of the work understudy focus on implementation efficiency, and flexibility is less attempted.We have explored the design space of Convolutional neural networks (CNNs) on Epiphany manycore architecture. We adopted a pipeline implementation of CNN that relies on the on-chip memory solely to store the weights. Also, the proposed mapping supported both ALexNet and GoogleNet CNN models, varying precision for weights, and two memory sizes for Epiphany cores. We were able to achieve competitive performance with respect to emerging manycores.As a part of the work in progress, we have studied a DL-architecture co-design approach to increase the flexibility of hardware solutions. A flexible platform should support variations in the model and variations in optimizations. The optimization method should be automated to respond to the changes in the model and application constraints with minor effort. Besides, the mapping of the models on embedded platforms should be automated as well.
|
|
