This thesis reports the work of using simulated annealing to design more efficient fuzzy logic systems to model problems with associated uncertainties. Simulated annealing is used within this work as a method for learning the best configurations of type-1 and type-2 fuzzy logic systems to maximise their modelling ability. Therefore, it presents the combination of simulated annealing with three models, type-1 fuzzy logic systems, interval type-2 fuzzy logic systems and general type-2 fuzzy logic systems to model four bench-mark problems including real-world problems. These problems are: noise-free Mackey-Glass time series forecasting, noisy Mackey-Glass time series forecasting and two real world problems which are: the estimation of the low voltage electrical line length in rural towns and the estimation of the medium voltage electrical line maintenance cost. The type-1 and type-2 fuzzy logic systems models are compared in their abilities to model uncertainties associated with these problems. Also, issues related to this combination between simulated annealing and fuzzy logic systems including type-2 fuzzy logic systems are discussed. The thesis contributes to knowledge by presenting novel contributions. The first is a novel approach to design interval type-2 fuzzy logic systems using the simulated annealing algorithm. Another novelty is related to the first automatic design of general type-2 fuzzy logic system using the vertical slice representation and a novel method to overcome some parametrisation difficulties when learning general type-2 fuzzy logic systems. The work shows that interval type-2 fuzzy logic systems added more abilities to modelling information and handling uncertainties than type-1 fuzzy logic systems but with a cost of more computations and time. For general type-2 fuzzy logic systems, the clear conclusion that learning the third dimension can add more abilities to modelling is an important advance in type-2 fuzzy logic systems research and should open the doors for more promising research and practical works on using general type-2 fuzzy logic systems to modelling applications despite the more computations associated with it.

Fuzzy logic systems are widely used for control, system identification, and pattern recognition problems. In order to maximize their performance, it is often necessary to undertake a design optimization process in which the adjustable parameters defining a particular fuzzy system are tuned to maximize a given performance criterion. Some data to approximate are commonly available and yield what is called the supervised learning problem. In this problem we typically wish to minimize the sum of the squares of errors in approximating the data. We first introduce fuzzy logic systems and the supervised learning problem that, in effect, is a nonlinear optimization problem that at times can be non-differentiable. We review the existing approaches and discuss their weaknesses and the issues involved. We then focus on one of these problems, i.e., non-differentiability of the objective function, and show how current approaches that do not account for non-differentiability can diverge. Moreover, we also show that non-differentiability may also have an adverse practical impact on algorithmic performances. We reformulate both the supervised learning problem and piecewise linear membership functions in order to obtain a polynomial or factorable optimization problem. We propose the application of a global nonconvex optimization approach, namely, a reformulation and linearization technique. The expanded problem dimensionality does not make this approach feasible at this time, even though this reformulation along with the proposed technique still bears a theoretical interest. Moreover, some future research directions are identified. We propose a novel approach to step-size selection in batch training. This approach uses a limited memory quadratic fit on past convergence data. Thus, it is similar to response surface methodologies, but it differs from them in the type of data that are used to fit the model, that is, already available data from the history of the algorithm are used instead of data obtained according to an experimental design. The step-size along the update direction (e.g., negative gradient or deflected negative gradient) is chosen according to a criterion of minimum distance from the vertex of the quadratic model. This approach rescales the complexity in the step-size selection from the order of the (large) number of training data, as in the case of exact line searches, to the order of the number of parameters (generally lower than the number of training data). The quadratic fit approach and a reduced variant are tested on some function approximation examples yielding distributions of the final mean square errors that are improved (i.e., skewed toward lower errors) with respect to the ones in the commonly used pattern-by-pattern approach. Moreover, the quadratic fit is also competitive and sometimes better than the batch training with optimal step-sizes, thus showing an improved performance of this approach. The quadratic fit approach is also tested in conjunction with gradient deflection strategies and memoryless variable metric methods, showing errors smaller by 1 to 7 orders of magnitude. Moreover, the convergence speed by using either the negative gradient direction or a deflected direction is higher than that of the pattern-by-pattern approach, although the computational cost of the algorithm per iteration is moderately higher than the one of the pattern-by-pattern method. Finally, some directions for future research are identified. / Ph. D.

No description available.

In the modern automotive industry, more and more manufacturers recognise that vehicle paint appearance makes an important contribution to customer satisfaction. Attractive appearance has become one of the important factors for customers in making a decision to purchase a car. Objective measurement of the quality of autobody paint appearance, as perceived by the customer, in a repeatable, reproducible, continuous scale manner is an important requirement for improving the paint appearance. It can provide car manufacturers a standard reference to evaluate the quality of the paint appearance. This thesis mainly deals with the measurement of paint appearance quality in the automotive industry by investigating, identifying and developing measurement methods in this area. First of all, the 'state of the art' in the area of paint appearance measurement was presented, which summarised the concept of appearance, models, attributes and definitions. To further identify the parameters and instruments used in the automotive industry, a round robin test was launched to perform visual assessment and instrument measurements on a set of panels in some European car manufacturers. A summary of the correlation found between measurable parameters and visual assessment provided the basis of the further work. Based on the literature survey and round robin test results, the next work is mainly concentrated on the two most important parameters, 'orange peel' and 'metal texture effect', how to separate and evaluate them. Digital signal processing technique, FFT and Filtering, have been employed to separate them and a set of measures have been provided for evaluation. At the same time, the technique for texture pattern recognition was introduced to evaluate the texture effect when a fine texture comparison was needed. A set of computable textural parameters based on grey-tone spatial-dependence matrices gives good correlation directly corresponding to visual perception. To resolve the overall appearance modelling problem, two novel and more powerful modelling tools, artificial neural networks and fuzzy logic, are introduced to model the overall appearance. The test results showed that both of them are able to reflect the correlation between overall appearance and the major parameters measured from a painted surface. Finally, an integrated measurement system, 'Smart Appearance', was developed using the image processing techniques and the artificial neural network model. The implement results show that this system can measure the major attributes of paint appearance and provide an overall appearance index corresponding to human visual perception. This system is helpful to product quality control on car body paint. It also could be used on the paint production line for dynamic measurement.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-03-10T12:31:18Z No. of bitstreams: 1 eduardopestanadeaguiar.pdf: 7884545 bytes, checksum: 182caace21281f7afce6554505811116 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-03-13T17:18:31Z (GMT) No. of bitstreams: 1 eduardopestanadeaguiar.pdf: 7884545 bytes, checksum: 182caace21281f7afce6554505811116 (MD5) / Made available in DSpace on 2017-03-13T17:18:31Z (GMT). No. of bitstreams: 1 eduardopestanadeaguiar.pdf: 7884545 bytes, checksum: 182caace21281f7afce6554505811116 (MD5) Previous issue date: 2016-09-26 / - / This thesis presents new fuzzy models applied to classification problems. With this regards, we introduce the use of set-membership concept, derived from the adaptive filter theory, into the training procedure of type-1 and singleton/non-singleton fuzzy logic systems, in order to reduce computational complexity and to increase convergence speed. Also, we present different criteria for using together with set-membership. Furthermore, we discuss the usefulness of delta rule delta, local Lipschitz estimation, variable step size and variable step size adaptive algorithms to yield additional improvement in terms of computational complexity reduction and convergence speed. Another important contribution of this thesis is to address the height type-reduction and to propose a modified version of interval singleton type-2 fuzzy logic system, so−called upper and lower singleton type-2 fuzzy logic system. The obtained results are compared with other models reported in the literature, demonstrating the effectiveness of the proposed classifiers and revealing that the proposals are able to properly handle with uncertainties associated with the measurements and with the data that are used to tune the parameters of the model. Based on data set provided by a Brazilian railway company, the models outlined above are applied in the classification of three possible faults and the normal condition of the switch machine, which is an equipment used for handling railroad switches. Finally, this thesis discusses the use of set-membership concept into the training procedure of an interval and singleton type-2 fuzzy logic system and of an upper and lower singleton type-2 fuzzy logic system, aiming to reduce computational complexity and to increase the convergence speed and the classification ratio. Also, we discuss the adoption of different criteria together with set-membership based-techniques. The performance is based on the data set composed of images provided by the same Brazilian railway company, which covers the four possible rail head defects and the normal condition of the rail head. The reported results show that the proposed models result in improved convergence speed, slightly higher classification ratio and remarkable computation complexity reduction when we limit the number of epochs for training, which may be required due to real time constraint or low computational resource availability.

Road transport mobility continues to be a challenge to the City of Johannesburg (CoJ)’s economy in general. Traffic signals, their remote monitoring and control systems are the current implemented Intelligent Transport Systems (ITS), but daily systematic faults and maintenance uncertainties on such systems decrease the effectiveness of traffic engineers’ intersections optimization techniques. Inefficient electrical power supply to such ITS is a challenge, with conditional power cuts and fluctuations, uncertainties on traffic control system faults. Another factor leading to the problem is the communication channel which is using traditional modems which are not reliable. Reporting through both customer complaints and such unreliable remote monitoring systems makes maintenance to be ineffective. In this dissertation, the factors leading to the faults and uncertainties are considered. The proposed solution considers the important concerns of ITS, such as electrical power source performance optimization technique, road traffic control systems compatibility and communications systems / Electrical and Mining Engineering / M. Tech. (Electrical Engineering)