教学课程资源库 更多>>

关键词： Failure   Success   Mathematical models   Decision making   Systems analysis   Nuclear accidents & safety   Probability   Algorithms   Nuclear power plants   Computer science   Mathematics  

摘要： Uncertainty quantification in modelling and simulation helps in identifying the factors that affect mechanics based and mathematical models. In recent years, there is an increased emphasis on formal quantification of risk under external hazards while also quantifying the associated uncertainties and their propagation. Probabilistic risk assessment (PRA) is a widely used tool by the nuclear industry for safety during normal operations as well as for protection against external hazards. Computation of total risk in an external hazard PRA is dependent on hazard assessment, fragility assessment, and systems analysis. Hence, uncertainty propagation in the assessment of external hazards requires a careful investigation in the context of systems analysis. A systems analysis for propagation of component fragilities is conducted using event and fault trees. The event and fault trees for an actual power plant can be fairly large in size, which imposes computational challenges. A fault and event tree analysis for a power plant requires modelling of hundreds of component failures, logic gates, multiple occurring events, and dependent events. Such interconnections create large networks that can lead to excessive computational demand. Hence, certain assumptions are employed for computational efficiency. Most of the traditional methods address computational demand by using assumptions or rely on high performance computing facilities that allow implementations of parallel computing. The assumptions typically represent the conditions imposed during the design basis (DB) scenarios. Traditional PRA tools based on these assumptions are also widely applied to perform risk assessment in the context of beyond design basis (BDB) scenarios. However, some of these assumptions may not be valid for certain BDB scenarios. In addition, the probability of dependent failures also increases in BDB scenarios due to common cause failures (CCF). The errors due to the assumptions in the traditional too

关键词： Computer science   Remote sensing   Technical communication   Information technology   Artificial intelligence  

摘要： The telecommunications industry is experiencing explosive data growth due to mobile devices' vast adoption. This scenario forces wireless operators to migrate to new efficient technologies like 5G to offer cost-efficient and scalable services to satisfy the consumers' demand. The management of the wireless network in this context requires new levels of automation and intelligence. The proposed anomaly detection algorithm for time series provides a solution to the autonomic vision of detecting outliers for time series in the context of wireless network operations while providing additional data cleaning capabilities. The algorithm was instantiated using Cloud technology in GCP and Confluent Cloud to verify its feasibility and performance in a real-world setting. The artifact was created following design science principles and guidelines. The evaluation results suggested that the proposed artifact provides anomaly detection and data cleaning performance similar to other state-of-the-art methods with the scalability required to support 5G and future technology generations in the wireless industry.

关键词： Neurosciences   Computer science  

关键词： Biochemistry   Biophysics   Computer science  

摘要： Biomolecules are dynamic entities that adopt a variety of conformations in solution. Conformational changes in biomolecules routinely take place when they take part in biochemical processes such as binding and catalysis. The energetic cost or conformational penalty to adopt an alternative conformation, which typically is paid for by thermal fluctuations or inter-molecular contacts with a partner molecule, can be an important determinant of these efficacy and selectivity of these biochemical processes. These conformational penalties can also be modulated by changes in physiological conditions and chemical modifications, enabling fine control over these biochemical processes, and aberrant changes to these conformational penalties can also be associated with disease. Thus, measurement of conformational penalties in biomolecules and how they can be tuned by external cues are essential to understand the role of conformational dynamics in biology. In this thesis, a combination of experimental and computational techniques such as NMR spectroscopy, UV melting and MD simulations are used to measure conformational penalties in nucleic acids and how they are modulated by post-transcriptional and epigenetic modifications, with particular applications to the formation of Watson-Crick like mismatches in DNA, and Hoogsteen base pairs in RNA and DNA. Improved methods that enable measurements of these conformational penalties with increased throughput and sensitivity, involving the use of NMR spectroscopy and UV melting, are also presented.

关键词： Neural networks   Computer science  

摘要： The modern software industry requires fast, highquality, and accurate forecasting of efforts and costs before the actual effort is invested in realizing the software product. Such requirements are a challenge for any software company, which must be ready to meet the expectations of the software customer. The main factor in the successful development of software projects and reducing the risk of errors is an adequate of the effort and costs invested during its implementation. In this doctoral dissertation, different approaches and models that have not been sufficiently precise and efficient so far will be analyzed, which resulted in only about 30% of successfully implemented software solutions. The main goal is to present three new, improved models based on efficient artificial intelligence tools, artificial neural networks. All three improved models use different architectures of artificial neural networks (ANN), constructed based on Taguchi's orthogonal vector plans. The goal is to optimize the improved models to avoid repeating the number of experiments and long time for their training. Applying the clustering method to several different sets of real projects further mitigates their heterogeneous structure. In addition, the input values of the projects are homogenized by the method of fuzzification, which achieves even greater reliability and accuracy of the obtained results. Optimization by the Taguchi method and increasing the coverage of a wide range of different projects leads to the efficient and successful completion of as many different software projects as possible. The main contributions of this paper are: constructing and identifying the best model for estimating effort and cost, selecting the best ANN architecture whose values converge the fastest to the minimum magnitude relative error, achieving a small number of experiments, reduced software effort estimation time due to convergence rate. Additional criteria and constraints are introduced to monitor

关键词： Computer science   Geographic information science   Artificial intelligence  

摘要： Location-aware technologies, such as personal navigation applications, location-based AR games, and artificial intelligence systems that learn from data about places, increasingly mediate our understanding of and interactions with the world. However, a number of risks associated with location-aware technologies have emerged, jeopardizing the welfare of its users. This dissertation seeks to understand and mitigate two such emerging risks associated with location-aware technologies - safety and geographic biases. The first part of this dissertation focuses on understanding and mitigating the safety issues in one of the most popular location-aware technologies - personal navigation systems. We analyzed catastrophic incidents that involved the use of personal navigation applications, identified technological causes that contributed to these incidents, and designed solutions to mitigate the most prominent technological causes. The second part of this dissertation focuses on mitigating the geographic biases in two types of location-aware technologies that leverage data-driven artificial intelligence - location-aware technologies based on computer vision and location-aware technologies based on knowledge base.

关键词： Computer science   Artificial intelligence  

关键词： Delays   Estimation   Bayes methods   Smart cards   Logic gates   Computer science   Clustering algorithms   AFC   automatic fare collection   AVL   automatic vehicle location   Bayes statistics   machine learning   missing data   smart card   train logs  

摘要： Smart card data from the Automatic Fare Collecting systems (AFC) and timetable information, such as Automatic Vehicle Location (AVL), are used in combination by practitioners and researchers to gain a deeper understanding of the public transit network. In some cases, AVL data are not available due to records being missing in the system. In such cases, people resort to the used schedule timetable such as General Transit Feed Specification (GTFS) to match smart card data to the transit network. Since delays or changes to the timetable are not contained in the scheduled timetable, it can result in wrong matches between the smart card data and the transit network. This paper shows how the uncertainty of arrival and departure times affects passengers to train assignments and proposes a method for estimating the missing arrival time of trains when the recorded timetable information is not available. The method uses the knowledge of how the tap-outs are distributed in a hierarchical, latent Bayesian model to predict the arrival times of trains. Evaluated on 15,136 train arrivals, the model can infer 70% of the arrivals times with an average error of 28 to 32 seconds depending on the station.

关键词： Computer engineering   Computer science   Artificial intelligence  

摘要： Deep neural networks (DNN) are proved to be effective and improve the performance dramatically in various kinds of computer vision tasks. The end-to-end learning manner in training DNN consistently shows the powerful modeling ability and consequently mitigates the dedicated efforts for expert feature engineering. On the other hand, it raises the issue that how to improve the black-box network with better representation (feature) learning especially when the learned representations and classifiers are tied together in the manner of supervised learning. In this work, representation learning is studied in four perspectives of different fields, i.e. diversity in ensemble learning, aspect ratio in image aesthetics assessment, invariance in identification task, and composition in color attribute recognition. In light of analyzing the bottleneck of black-box network and designing better representation learning for target tasks, we introduce that: (a) Ensemble learning relies on the diversity of the complementary neural networks, in both feature representations and classifier representations. A diverse representation learning method, namely learning-difficulty-aware embedding, is proposed to adaptively reconcile learning attentions for different categories by training a series of networks with diversified representations sequentially; (b) Widely-adopted data augmentation in image recognition deteriorates aspect ratios, which is an important factor in image aesthetics assessment. An aspect ratio representation learning method, namely adaptive fractional dilated convolution, is proposed to explicitly preserve the learning representation related to aspect ratios by adjusting the receptive fields adaptively and natively; (c) Identification tasks, e.g. person re-identification, aim at learning representations that are robust to interfering variances, e.g. lighting variances, view variances, pose variances. An invariance representation learning method, namely anchor loss, is prop

关键词： Fluid mechanics   Computer science   Materials science   Mathematics   Mechanics   Physics  

摘要： Joining of dissimilar metals is a process that is of interest in many fields, especially the automotive industry where lightweighting of the body structure is important. However, creating strong joints between dissimilar metals can be challenging. Friction bit joining (FBJ) is a solid-state method that uses a consumable bit to create a strong joint between dissimilar metals such as aluminum and steel. The purpose of this research is to gain understanding of how adjusting FBJ parameters affects the heating and pressure at the weld interface using a modeling approach, in order to better understand the bonding process. The questions guiding this research are: (1) What is the effect of spindle speed, plunge rate, and plunge depth on joint strength? (2) Can the proposed model be developed with enough fidelity to correlate the effect of these parameters on joint strength, within 10%? (3) What is the effect of the simulated vertical load profile on heating at the interface? (4) Does the load profile/heating relationship correlate to experimental joint strength to within 10%? A design of experiments approach found that the effect of spindle speed on joint strength is significant. Plunge rate did not have a significant effect, but the interaction between plunge rate and spindle speed was significant. A model was created, and multiple simulations were run to study these interactions. Initial simulations were run based on the input parameters used for the experiments. The simulation data was used to run a full second order regression was run which found that spindle speed had a significant effect on the experimental Z load. The data also revealed that spindle speed and plunge rate have a strong correlation between bonded area and temperature. Simulated versus experimental Z loads have a good correlation. Experimental bonded area had a slight correlation to joint strength trending in the correct direction. The shape of the simulated cross section did not fully match the experim