arXiv subjects

T. Marwala

Publications and source records attributed to T. Marwala.

At least 19 recordsLinked to original sources

arXiv2017

A Differential Evaluation Markov Chain Monte Carlo algorithm for Bayesian Model Updating

The use of the Bayesian tools in system identification and model updating paradigms has been increased in the last ten years. Usually, the Bayesian techniques can be implemented to incorporate the uncertainties associated with measurements as well as the prediction made by the finite element model (FEM) into the FEM updating procedure. In this case, the posterior distribution function describes the uncertainty in the FE model prediction and the experimental data. Due to the complexity of the modeled systems, the analytical solution for the posterior distribution function may not exist. This leads to the use of numerical methods, such as Markov Chain Monte Carlo techniques, to obtain approximate solutions for the posterior distribution function. In this paper, a Differential Evaluation Markov Chain Monte Carlo (DE-MC) method is used to approximate the posterior function and update FEMs. The main idea of the DE-MC approach is to combine the Differential Evolution, which is an effective global optimization algorithm over real parameter space, with Markov Chain Monte Carlo (MCMC) techniques to generate samples from the posterior distribution function. In this paper, the DE-MC method is discussed in detail while the performance and the accuracy of this algorithm are investigated by updating two structural examples.

cs.CE↗

arXiv2017

Fuzzy finite element model updating using metaheuristic optimization algorithms

In this paper, a non-probabilistic method based on fuzzy logic is used to update finite element models (FEMs). Model updating techniques use the measured data to improve the accuracy of numerical models of structures. However, the measured data are contaminated with experimental noise and the models are inaccurate due to randomness in the parameters. This kind of aleatory uncertainty is irreducible, and may decrease the accuracy of the finite element model updating process. However, uncertainty quantification methods can be used to identify the uncertainty in the updating parameters. In this paper, the uncertainties associated with the modal parameters are defined as fuzzy membership functions, while the model updating procedure is defined as an optimization problem at each α-cut level. To determine the membership functions of the updated parameters, an objective function is defined and minimized using two metaheuristic optimization algorithms: ant colony optimization (ACO) and particle swarm optimization (PSO). A structural example is used to investigate the accuracy of the fuzzy model updating strategy using the PSO and ACO algorithms. Furthermore, the results obtained by the fuzzy finite element model updating are compared with the Bayesian model updating results.

cs.AI↗

arXiv2014

Missing Data Prediction and Classification: The Use of Auto-Associative Neural Networks and Optimization Algorithms

This paper presents methods which are aimed at finding approximations to missing data in a dataset by using optimization algorithms to optimize the network parameters after which prediction and classification tasks can be performed. The optimization methods that are considered are genetic algorithm (GA), simulated annealing (SA), particle swarm optimization (PSO), random forest (RF) and negative selection (NS) and these methods are individually used in combination with auto-associative neural networks (AANN) for missing data estimation and the results obtained are compared. The methods suggested use the optimization algorithms to minimize an error function derived from training the auto-associative neural network during which the interrelationships between the inputs and the outputs are obtained and stored in the weights connecting the different layers of the network. The error function is expressed as the square of the difference between the actual observations and predicted values from an auto-associative neural network. In the event of missing data, all the values of the actual observations are not known hence, the error function is decomposed to depend on the known and unknown variable values. Multi-layer perceptron (MLP) neural network is employed to train the neural networks using the scaled conjugate gradient (SCG) method. Prediction accuracy is determined by mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and correlation coefficient (r) computations. Accuracy in classification is obtained by plotting ROC curves and calculating the areas under these. Analysis of results depicts that the approach using RF with AANN produces the most accurate predictions and classifications while on the other end of the scale is the approach which entails using NS with AANN.

cs.NE↗

arXiv2013

Finite Element Model Updating Using Fish School Search Optimization Method

A recent nature inspired optimization algorithm, Fish School Search (FSS) is applied to the finite element model (FEM) updating problem. This method is tested on a GARTEUR SM-AG19 aeroplane structure. The results of this algorithm are compared with two other metaheuristic algorithms; Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). It is observed that on average, the FSS and PSO algorithms give more accurate results than the GA. A minor modification to the FSS is proposed. This modification improves the performance of FSS on the FEM updating problem which has a constrained search space.

cs.CE↗

arXiv2012

Common Mistakes when Applying Computational Intelligence and Machine Learning to Stock Market modelling

For a number of reasons, computational intelligence and machine learning methods have been largely dismissed by the professional community. The reasons for this are numerous and varied, but inevitably amongst the reasons given is that the systems designed often do not perform as expected by their designers. The reasons for this lack of performance is a direct result of mistakes that are commonly seen in market-prediction systems. This paper examines some of the more common mistakes, namely dataset insufficiency; inappropriate scaling; time-series tracking; inappropriate target quantification and inappropriate measures of performance. The rationale that leads to each of these mistakes is examined, as well as the nature of the errors they introduce to the analysis / design. Alternative ways of performing each task are also recommended in order to avoid perpetuating these mistakes, and hopefully to aid in clearing the way for the use of these powerful techniques in industry.

stat.AP↗

arXiv2011

Sampling Techniques in Bayesian Finite Element Model Updating

Recent papers in the field of Finite Element Model (FEM) updating have highlighted the benefits of Bayesian techniques. The Bayesian approaches are designed to deal with the uncertainties associated with complex systems, which is the main problem in the development and updating of FEMs. This paper highlights the complexities and challenges of implementing any Bayesian method when the analysis involves a complicated structural dynamic model. In such systems an analytical Bayesian formulation might not be available in an analytic form; therefore this leads to the use of numerical methods, i.e. sampling methods. The main challenge then is to determine an efficient sampling of the model parameter space. In this paper, three sampling techniques, the Metropolis-Hastings (MH) algorithm, Slice Sampling and the Hybrid Monte Carlo (HMC) technique, are tested by updating a structural beam model. The efficiency and limitations of each technique is investigated when the FEM updating problem is implemented using the Bayesian Approach. Both MH and HMC techniques are found to perform better than the Slice sampling when Young's modulus is chosen as the updating parameter. The HMC method gives better results than MH and Slice sampling techniques, when the area moment of inertias and section areas are updated.

cs.CE↗

arXiv2011

Single-trial EEG Discrimination between Wrist and Finger Movement Imagery and Execution in a Sensorimotor BCI

A brain-computer interface (BCI) may be used to control a prosthetic or orthotic hand using neural activity from the brain. The core of this sensorimotor BCI lies in the interpretation of the neural information extracted from electroencephalogram (EEG). It is desired to improve on the interpretation of EEG to allow people with neuromuscular disorders to perform daily activities. This paper investigates the possibility of discriminating between the EEG associated with wrist and finger movements. The EEG was recorded from test subjects as they executed and imagined five essential hand movements using both hands. Independent component analysis (ICA) and time-frequency techniques were used to extract spectral features based on event-related (de)synchronisation (ERD/ERS), while the Bhattacharyya distance (BD) was used for feature reduction. Mahalanobis distance (MD) clustering and artificial neural networks (ANN) were used as classifiers and obtained average accuracies of 65 % and 71 % respectively. This shows that EEG discrimination between wrist and finger movements is possible. The research introduces a new combination of motor tasks to BCI research.

cs.AI↗

arXiv2011

Ant Colony Optimization of Rough Set for HV Bushings Fault Detection

Most transformer failures are attributed to bushings failures. Hence it is necessary to monitor the condition of bushings. In this paper three methods are developed to monitor the condition of oil filled bushing. Multi-layer perceptron (MLP), Radial basis function (RBF) and Rough Set (RS) models are developed and combined through majority voting to form a committee. The MLP performs better that the RBF and the RS is terms of classification accuracy. The RBF is the fasted to train. The committee performs better than the individual models. The diversity of models is measured to evaluate their similarity when used in the committee.

cs.NE↗

arXiv2011

Artificial Neural Network and Rough Set for HV Bushings Condition Monitoring

cs.NE↗

arXiv2008

Missing Data using Decision Forest and Computational Intelligence

Autoencoder neural network is implemented to estimate the missing data. Genetic algorithm is implemented for network optimization and estimating the missing data. Missing data is treated as Missing At Random mechanism by implementing maximum likelihood algorithm. The network performance is determined by calculating the mean square error of the network prediction. The network is further optimized by implementing Decision Forest. The impact of missing data is then investigated and decision forrests are found to improve the results.

stat.ML↗

arXiv2008

Land Cover Mapping Using Ensemble Feature Selection Methods

Ensemble classification is an emerging approach to land cover mapping whereby the final classification output is a result of a consensus of classifiers. Intuitively, an ensemble system should consist of base classifiers which are diverse i.e. classifiers whose decision boundaries err differently. In this paper ensemble feature selection is used to impose diversity in ensembles. The features of the constituent base classifiers for each ensemble were created through an exhaustive search algorithm using different separability indices. For each ensemble, the classification accuracy was derived as well as a diversity measure purported to give a measure of the inensemble diversity. The correlation between ensemble classification accuracy and diversity measure was determined to establish the interplay between the two variables. From the findings of this paper, diversity measures as currently formulated do not provide an adequate means upon which to constitute ensembles for land cover mapping.

cs.LG↗

arXiv2008

Relationship between Diversity and Perfomance of Multiple Classifiers for Decision Support

The paper presents the investigation and implementation of the relationship between diversity and the performance of multiple classifiers on classification accuracy. The study is critical as to build classifiers that are strong and can generalize better. The parameters of the neural network within the committee were varied to induce diversity; hence structural diversity is the focus for this study. The hidden nodes and the activation function are the parameters that were varied. The diversity measures that were adopted from ecology such as Shannon and Simpson were used to quantify diversity. Genetic algorithm is used to find the optimal ensemble by using the accuracy as the cost function. The results observed shows that there is a relationship between structural diversity and accuracy. It is observed that the classification accuracy of an ensemble increases as the diversity increases. There was an increase of 3%-6% in the classification accuracy.

cs.AI↗

arXiv2008

The use of entropy to measure structural diversity

In this paper entropy based methods are compared and used to measure structural diversity of an ensemble of 21 classifiers. This measure is mostly applied in ecology, whereby species counts are used as a measure of diversity. The measures used were Shannon entropy, Simpsons and the Berger Parker diversity indexes. As the diversity indexes increased so did the accuracy of the ensemble. An ensemble dominated by classifiers with the same structure produced poor accuracy. Uncertainty rule from information theory was also used to further define diversity. Genetic algorithms were used to find the optimal ensemble by using the diversity indices as the cost function. The method of voting was used to aggregate the decisions.

cs.LG↗

arXiv2007

On-Line Condition Monitoring using Computational Intelligence

This paper presents bushing condition monitoring frameworks that use multi-layer perceptrons (MLP), radial basis functions (RBF) and support vector machines (SVM) classifiers. The first level of the framework determines if the bushing is faulty or not while the second level determines the type of fault. The diagnostic gases in the bushings are analyzed using the dissolve gas analysis. MLP gives superior performance in terms of accuracy and training time than SVM and RBF. In addition, an on-line bushing condition monitoring approach, which is able to adapt to newly acquired data are introduced. This approach is able to accommodate new classes that are introduced by incoming data and is implemented using an incremental learning algorithm that uses MLP. The testing results improved from 67.5% to 95.8% as new data were introduced and the testing results improved from 60% to 95.3% as new conditions were introduced. On average the confidence value of the framework on its decision was 0.92.

cs.AI↗

arXiv2007

Response Prediction of Structural System Subject to Earthquake Motions using Artificial Neural Network

This paper uses Artificial Neural Network (ANN) models to compute response of structural system subject to Indian earthquakes at Chamoli and Uttarkashi ground motion data. The system is first trained for a single real earthquake data. The trained ANN architecture is then used to simulate earthquakes with various intensities and it was found that the predicted responses given by ANN model are accurate for practical purposes. When the ANN is trained by a part of the ground motion data, it can also identify the responses of the structural system well. In this way the safeness of the structural systems may be predicted in case of future earthquakes without waiting for the earthquake to occur for the lessons. Time period and the corresponding maximum response of the building for an earthquake has been evaluated, which is again trained to predict the maximum response of the building at different time periods. The trained time period versus maximum response ANN model is also tested for real earthquake data of other place, which was not used in the training and was found to be in good agreement.

cs.AI↗

arXiv2007

Scalability and Optimisation of a Committee of Agents Using Genetic Algorithm

A population of committees of agents that learn by using neural networks is implemented to simulate the stock market. Each committee of agents, which is regarded as a player in a game, is optimised by continually adapting the architecture of the agents using genetic algorithms. The committees of agents buy and sell stocks by following this procedure: (1) obtain the current price of stocks; (2) predict the future price of stocks; (3) and for a given price trade until all the players are mutually satisfied. The trading of stocks is conducted by following these rules: (1) if a player expects an increase in price then it tries to buy the stock; (2) else if it expects a drop in the price, it sells the stock; (3)and the order in which a player participates in the game is random. The proposed procedure is implemented to simulate trading of three stocks, namely, the Dow Jones, the Nasdaq and the S&P 500. A linear relationship between the number of players and agents versus the computational time to run the complete simulation is observed. It is also found that no player has a monopolistic advantage.

cs.MA↗

arXiv2007

Principal Component Analysis and Automatic Relevance Determination in Damage Identification

This paper compares two neural network input selection schemes, the Principal Component Analysis (PCA) and the Automatic Relevance Determination (ARD) based on Mac-Kay's evidence framework. The PCA takes all the input data and projects it onto a lower dimension space, thereby reduc-ing the dimension of the input space. This input reduction method often results with parameters that have significant influence on the dynamics of the data being diluted by those that do not influence the dynamics of the data. The ARD selects the most relevant input parameters and discards those that do not contribute significantly to the dynamics of the data being modelled. The ARD sometimes results with important input parameters being discarded thereby compromising the dynamics of the data. The PCA and ARD methods are implemented together with a Multi-Layer-Perceptron (MLP) network for fault identification in structures and the performance of the two methods is as-sessed. It is observed that ARD and PCA give similar accu-racy levels when used as input-selection schemes. There-fore, the choice of input-selection scheme is dependent on the nature of the data being processed.

cs.CE↗

arXiv2007

Using artificial intelligence for data reduction in mechanical engineering

In this paper artificial neural networks and support vector machines are used to reduce the amount of vibration data that is required to estimate the Time Domain Average of a gear vibration signal. Two models for estimating the time domain average of a gear vibration signal are proposed. The models are tested on data from an accelerated gear life test rig. Experimental results indicate that the required data for calculating the Time Domain Average of a gear vibration signal can be reduced by up to 75% when the proposed models are implemented.

cs.CE↗