Local active noise control systems generate a zone of quiet at the physical error sensor location. While significant attenuation is achieved at the error sensor, local noise control is not without its problems, chiefly that the zone of quiet is generally small and impractically sized. It may be inconvenient to place the error sensor at the desired location of attenuation, such as near an observer’s ear, preventing the small zone of quiet from being centered there. To overcome the problems encountered in local active noise control, virtual acoustic sensors have been developed to shift the zone of quiet away from the physical sensor position to a spatially fixed desired location. The general aim of the research presented in this thesis is to improve and extend the spatially fixed and moving virtual sensing algorithms developed for active noise control thus far and hence increase the scope and application of local active noise control systems. To achieve this research aim, a number of novel spatially fixed and moving virtual sensing algorithms are presented for local active noise control. In this thesis, a spatially fixed virtual sensing technique named the Stochastically Optimal Tonal Diffuse Field (SOTDF) virtual sensing method is developed specifically for use in pure tone diffuse sound fields. The SOTDF virtual sensing method is a fixed gain virtual sensing method that does not require a preliminary identification stage nor models of the complex transfer functions between the error sensors and the sources. SOTDF virtual microphones and virtual energy density sensors that use both pressure and pressure gradient sensors are developed using the SOTDF virtual sensing method. The performance of these SOTDF virtual sensors is investigated in numerical simulations and using experimental measurements made in a reverberation chamber. SOTDF virtual sensors are shown to accurately estimate the pressure and pressure gradient at a virtual location and to effectively shift the zone of quiet away from the physical sensors to the virtual location. In numerically simulated and post-processed experimental control, both virtual microphones and virtual energy density sensors achieve higher attenuation at the virtual location than conventional control strategies employing their physical counterpart. As it is likely that the desired location of attenuation is not spatially fixed, a number of moving virtual sensing algorithms are also developed in this thesis. These moving virtual sensing algorithms generate a virtual microphone that tracks the desired location of attenuation as it moves through a three-dimensional sound field. To determine the level of attenuation that can be expected at the ear of a seated observer in practice, the performance of the moving virtual sensing algorithms in generating a moving zone of quiet at the single ear of a rotating artificial head is investigated in real-time experiments conducted in a modally dense three dimensional cavity. Results of real-time experiments demonstrate that moving virtual sensors provide improved attenuation at the moving virtual location compared to either fixed virtual sensors or fixed physical sensors. As an acoustic energy density cost function spatially extends the zone of quiet generated at the sensor location, a fixed three-dimensional virtual acoustic energy density sensing method is also developed for use in a modally dense three-dimensional sound field. The size of the localised zone of quiet achieved by minimising either the acoustic energy density or the squared pressure at the virtual location with the active noise control system is compared in real-time experiments conduced in a modally dense three-dimensional cavity. Experimental results demonstrate that minimising the virtual acoustic energy density provides improved attenuation in the sound field and a larger 10 dB zone of quiet at the virtual location than virtual microphones. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1522526 / Thesis (Ph.D.) -- University of Adelaide, School of Mechanical Engineering, 2010

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第21917号 / 情博第700号 / 新制||情||120(附属図書館) / 京都大学大学院情報学研究科システム科学専攻 / (主査)教授 加納 学, 教授 杉江 俊治, 教授 大塚 敏之 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

This dissertation explores the problem of energy estimation in supervised Non-Intrusive Load Monitoring (NILM). NILM refers to a set of techniques used to estimate the electricity consumed by individual loads in a building from measurements of the total electrical consumption. Most commonly, NILM works by first attributing any significant change in the total power consumption (also known as an event) to a specific load and subsequently using these attributions (i.e. the labels for the events) to estimate energy for each load. For this last step, most proposed solutions in the field impart simplifying assumptions to make the problem more tractable. This has severely limited the practicality of the proposed solutions. To address this knowledge gap, we present a framework for creating appliance models based on classification labels and aggregate power measurements that can help relax many of these assumptions. Within the framework, we model the problem of utilizing a sequence of event labels to generate energy estimates as a broader class of problems that has two major components (i) With the understanding that the labels arise from a process with distinct states and state transitions, we estimate the underlying Finite State Machine (FSM) model that most likely generated the observed sequence (ii) We allow for the observed sequence to have errors, and present an error correction algorithm to detect and correct them. We test the framework on data from 43 appliances collected from 19 houses and find that it improves errors in energy estimates when compared to the case with no correction in 19 appliances by a factor of 50, leaves 17 appliances unchanged, and negatively impacts 6 appliances by a factor of 1.4. This approach of utilizing event sequences to estimate energy has implications in virtual metering of appliances as well. In a case study, we utilize this framework in order to substitute the need of plug-level sensors with cheap and easily deployable contacless sensors, and find that on the 6 appliances virtually metered using magnetic field sensors, the inferred energy values have an average error of 10:9%.

In our thesis we explore the Automatic Question/Answer Generation (AQAG) and the application of Machine Learning (ML) in natural language queries. Initially we create a collection of question/answer tuples conceptually based on processing received data from (virtual) sensors placed in a smart city. Subsequently we train a Gated Recurrent Unit(GRU) model on the generated dataset and evaluate the accuracy we can achieve in answering those questions. This will help in turn to address the problem of automatic sensor composition based on natural language queries. To this end, the contribution of this thesis is two-fold: on one hand we are providing anautomatic procedure for dataset construction, based on natural language question templates, and on the other hand we apply a ML approach that establishes the correlation between the natural language queries and their virtual sensor representation, via their functional representation. We consider virtual sensors to be entities as described by Mihailescu et al, where they provide an interface constructed with certain properties in mind. We use those sensors for our application domain of a smart city environment, thus constructing our dataset around questions relevant to it.

This thesis treats virtual sensing for fatigue assessment of steel bridges. The purpose is to develop avirtual sensing method to use in the fatigue assessment process. The aim for the virtual sensing method is to only depend on strain measurements located on the bridge structure. The service life of bridges is often limited by fatigue and amending bridge design to improve fatigue resistance was developed in the 1970s. There are several bridges in Sweden, Europe and other countries that have exceeded their theoretical service life with regard to fatigue, and the need to replace them isboth a environmental and economical issue. The bridge over Rautasjokk north of Kiruna, Sweden is a specific example where the theoretical service life is limited by fatigue. Uncertainties in the theoretical fatigue assessment of bridges can be reduced by measuring strains atthe fatigue critical details, and therefore lead to a longer theoretical service life. Monitoring is, however,an expensive method and the procedure of installation and administration requires working time, and monitoring can only provide information at the gauge location. Hence, it is of great interest to optimizethe monitoring system. Virtual sensing is a method that could provide an alternative to conventionalmeasuring techniques. Virtual sensing combine measurement data with information from a model. Virtual sensing for fatigue assessment of the Rautasjokk Bridge was evaluated developing two methods. Both methods uses a finite element model of the bridge combined with strain measurements. The measurements were obtained on February 14 2018 and included a time signal and strain variations at six different locations of the structure. The accuracy of the virtual sensing methods were evaluated by comparing the fatigue damage of virtual sensing with the fatigue damage calculated using measured strains. The fatigue calculations were based on methods presented in the Eurocode EN-1993-1-9. The first method was based on the idea to find a relation between groups of stress ranges for two gauge locations on the bridge. The stress ranges were established by loading influence lines obtained from the finite element model with a fictitious train and the difference between two gauges was stored in a vector, the correlation vector. The correlation vector was applied on the measured stress ranges of the first gauge to estimate the actual stress ranges of the second gauge. No relation between groups of stress ranges for different loading cases was found and the correlation vector method for virtual sensing is not a sufficiently accurate method to apply in the fatigue assessment of the Rautasjokk Bridge. The second method was based on finding a relation between each stress range instead of a group of stress ranges. Influence lines from the finite element model were used to find a relation between each stress range of two gauges. Their relation was stored in a matrix, the correlation matrix. The matrix was applied on the measured stress ranges of the first gauge to estimate the actual stress ranges of the second gauge. The correlation matrix method for virtual sensing estimate the fatigue damage sufficiently accurate at the bridge locations where local stress ranges have the greatest impact on the fatigue damage. Results obtained through virtual sensing only include the same parameters that were used as inputs in the method. A credible virtual sensing method is crucial in order to achieve reliable results. In general, a virtual sensing method requires an extent amount of input data to validate its reliability. Further studies are required to investigate how the uncertainties of the correlation matrix affect the fatigue assessment.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第18310号 / 工博第3902号 / 新制||工||1598(附属図書館) / 31168 / 京都大学大学院工学研究科化学工学専攻 / (主査)教授 長谷部 伸治, 教授 大嶋 正裕, 教授 河瀬 元明 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

A indústria automobilística é um dos mais importantes setores da economia no Brasil e no mundo. Nos últimos anos viu-se praticamente obrigada a melhorar o desempenho de seus veículos produzidos e reduzir seus custos. Um dos marcos desta transformação foi o desenvolvimento do sensor de oxigênio, sendo este um dos principais elementos dos sistemas gerenciadores de motor. Esta dissertação propõe o estudo de arquiteturas de sistemas inteligentes para sensoriamento virtual de oxigênio em veículos bicombustíveis, utilizando-se redes neurais artificiais supervisionadas, com arquitetura Perceptron multicamadas. As topologias implementadas atingiram resultados com erros relativos médios menores que 1% em centenas de topologias. Verificou-se também que para o sensoriamento virtual de oxigênio em veículos bicombustíveis, a abordagem de se realizar treinamentos com todos os tipos de combustíveis, segmentando conjuntos de todo o universo de dados, mostra-se a mais adequada. / The automotive industry is one of the most important sectors in Brazilians economy and in the world. In recent years, this industry has been forced to improve the performance of their produced vehicles and to reduce their costs. One of the landmarks of this transformation was the development of the oxygen sensor, which is one of the main elements of the engine management systems. This dissertation proposes the use of intelligent systems architectures for virtual oxygen sensing of bi-fuel vehicles, using multilayer Perceptron artificial neural networks. The implemented topologies reach results with mean relative errors less than 1% in hundreds of topologies. It was also noted that the approach to train the neural network with all types of fuels, using subsets of data universe, it is the most appropriate to have a virtual sensing of oxygen in bi-fuel vehicles.

A indústria automobilística é um dos mais importantes setores da economia no Brasil e no mundo. Nos últimos anos viu-se praticamente obrigada a melhorar o desempenho de seus veículos produzidos e reduzir seus custos. Um dos marcos desta transformação foi o desenvolvimento do sensor de oxigênio, sendo este um dos principais elementos dos sistemas gerenciadores de motor. Esta dissertação propõe o estudo de arquiteturas de sistemas inteligentes para sensoriamento virtual de oxigênio em veículos bicombustíveis, utilizando-se redes neurais artificiais supervisionadas, com arquitetura Perceptron multicamadas. As topologias implementadas atingiram resultados com erros relativos médios menores que 1% em centenas de topologias. Verificou-se também que para o sensoriamento virtual de oxigênio em veículos bicombustíveis, a abordagem de se realizar treinamentos com todos os tipos de combustíveis, segmentando conjuntos de todo o universo de dados, mostra-se a mais adequada. / The automotive industry is one of the most important sectors in Brazilians economy and in the world. In recent years, this industry has been forced to improve the performance of their produced vehicles and to reduce their costs. One of the landmarks of this transformation was the development of the oxygen sensor, which is one of the main elements of the engine management systems. This dissertation proposes the use of intelligent systems architectures for virtual oxygen sensing of bi-fuel vehicles, using multilayer Perceptron artificial neural networks. The implemented topologies reach results with mean relative errors less than 1% in hundreds of topologies. It was also noted that the approach to train the neural network with all types of fuels, using subsets of data universe, it is the most appropriate to have a virtual sensing of oxygen in bi-fuel vehicles.