Search

MEng (Electrical and Electronic Engineering), North-West University, Potchefstroom Campus Medium voltage cable terminations are critical components within the electricity network and are expected to have high reliability to ensure network performance and compliance to regulatory requirements. It is evident that failure rates are currently alarmingly high and indicates that more than 20% of some cable system failure relates to cable terminations. It appears that poor workmanship is the main contributing factor. The termination product itself is of high quality with well-established manufacturing methods and processes. The electricity industry is also well recognized with a vast number of different testing standards and methods for cable systems. Although good testing standards are by no means a guarantee that good workmanship was executed on site, it is a valuable method to verify the quality of the workmanship. This dissertation focuses on using online partial discharge measurements to optimize asset management of medium voltage cable terminations. Cable terminations as an asset portfolio within asset management is discussed with reference to international asset management standard. The importance of quality assurance is addressed by identifying current shortcomings and concerns. The dissertation is supported by making use of three methods. Termination modeling and simulation are done with a Python program by implementing the 3-capacitor model. Finite element analysis is done with QuickField software program. Actual partial discharge measurement data was obtained by conducting an accelerated aging experiment by creating five artificial termination defects. The last method gathered data from role-players within the industry by means of an online survey in the form of a questionnaire. The dissertation converges into the optimization of asset management for cable terminations by addressing suitable structures with the overall objective of quality assurance and reliable operations. A recommendation is made by formulating two new test methods for cable terminations by using unconventional partial discharge measurements. Masters

MEng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus This dissertation was created in pursuit of determining the viability of a smartphone based technique for metabolic syndrome screening. This screening method consists of two artificial neural networks. The first network uses a photoplethysmography (PPG) waveform measured using a smartphone camera and its FFT analysis to determine blood pressure. The second classifier uses the heart rate and blood pressure determined from the PPG measurement and user data, including sex, age, BMI, waist-to-height ratio, smoking, drinking habits, medical history and activity level, to determine whether a user may be at risk of metabolic syndrome. The blood pressure network is initially trained and tested using PPG data, measured with a blood oximeter, from the MIMIC III online waveform dataset. The algorithm is then tested with PPG data obtained from participants with a smartphone camera. In real world scenarios the BP (blood pressure) algorithm has a mean absolute error of 7.7 mmHg (±6:9) for SBP (systolic blood pressure) and 8.0 mmHg (±6:3) for DBP (diastolic blood pressure). Multiple metabolic syndrome algorithms are tested, with and without lifestyle related factors, and with a simulated error being added to blood pressure inputs based on results from participant tests. It was found that the inclusion of lifestyle factors improved the performance of the model by approximately 2%, from 75.2% to 77.2%, when assessed using 10-fold cross validation. When including simulated HR and BP errors, the network including lifestyle factors achieved an average accuracy of 74.9%. An approach similar to bootstrapping was followed where the 10-cross validation process is executed 10 times, randomly shuffling the data with each cycle, resulting in a total of 100 testing sets. The average accuracy across the 100 sets for the first model was 73.4%, with a sensitivity of 73.1% and a specificity of 73.6%. For the model with lifestyle factors the accuracy was 76.3%, with a sensitivity of 74.9% and a specificity of 77.4%. A t-test was done to show that the accuracy difference between the models was statistically significant (t = -2.48; p < 0.014). A bias in the BP algorithm was found for the real world testing, causing the mean SBP to be 5 mmHg too low. Due to a strong dependence on DBP, this caused the MetS algorithm to perform less optimally. This highlights the necessity of an accurate BP algorithm for the MetS algorithm to provide reliable results. The accuracy obtained across the 100 testing sets when including HR and BP errors was 73.1%, with a sensitivity of 64.9% and a specificity of 79.9%. In order to determine if the output from the model could be interpreted as a risk indicator instead of just a binary prediction, the models were assessed at different confidence levels. The models achieved 86.8% to 88.4% accuracy for confidence levels over 80% and 92.7% to 94.1% accuracy for confidence over 90%. This seems to indicate that there is definite value in using this model as a risk indicator, which a user could potentially use to determine whether a full diagnosis is warranted. In that sense, it falls perfectly in line with what is expected from a screening solution. The results show that smartphone based metabolic syndrome screening is a promising venture, that can be expanded upon in further research and considered for real world screening solutions. However, further research would be required to test whether this solution achieves similar results across multiple devices before it could be rolled out as a final product. Masters

MEng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus In portfolio management, asset allocation is one of the most crucial and difficult challenges investors face. Asset allocation is defined as a decision making process of spreading available funds into various financial assets. The most famous and widely used models for tackling asset allocation problems are mean variance, mean valueatrisk and Sharpe ratio. These models are solvable by quadratic programming, and they all rely heavily on the mean and standard deviation with the assumption that the data distribution is symmetrical. Unfortunately, a majority of the realworld problems exhibit asymmetric distributions; as a result, the modified Sharpe ratio is introduced to include skewness and kurtosis as the third and fourth moments of return. The results obtained in this study are based on the modified Sharpe ratio, and they apply and compare genetic algorithm, particle swarm optimisation, and deep deterministic policy gradient to solve the asset allocation problem. The former algorithms (genetic algorithms and particle swarm optimisation) are widely employed to generate high quality solutions in optimisation problems whilst the latter (deep deterministic policy gradient) has proved to be more effective in solving complex problems that cannot be solved by conventional techniques. The algorithms learn to evolve portfolio weights in maximising the modified Sharpe ratio. The dataset used is extracted from the banking sector of the Johannesburg stock exchange and wellknown stocks in the United States stock exchange. In measuring the performance of the three algorithms, a uniform allocation is used as a baseline asset allocation strategy. Uniform allocation divides portfolio weights equally among the assets in a portfolio. The results presented show that all three algorithms outclass the uniform allocation on numerous occasions. In general, the genetic algorithm and particle swarm optimisation provide relatively better results than deep deterministic policy gradient. The results are then tested on buyandhold. Even though the deep deterministic policy gradient did not perform well in evolving portfolio weights and took too long to run in training, it is comparable with the uniform allocation. The genetic algorithm outperforms the other algorithms with particle swarm optimisation following. Masters

Condition monitoring of gearboxes is a crucial task because gearboxes are essential power transmission components whose failure can lead to a catastrophic breakdown of machines. The common faults of a gearbox system, such as tooth breakage, wear, scuffing, spalling and lubricant starvation, have a significant influence on the inside friction and heat dissipation, and consequently, it changes the temperature field distribution within the gearbox. Thermal imaging is a promising technique in the field of machine condition monitoring via the variation detection of heat distribution. However, the thermal images require significant storage space, a high transfer rate and high-speed hardware. To achieve intelligent and efficient machine condition monitoring with the advanced thermal imaging technique, this study reduces the dimensionality of thermal images of a two-stage gearbox system via compressive sensing (CS) and classifies three different lubricant shortage conditions based on the compressed features with an intelligent convolutional neural network (CNN). The experimental results demonstrate that the compressed thermal images contain sufficient fault information and are capable of diagnosing the inadequate lubrication faults for gearboxes operating at various working conditions.

Backup bearings (BBs) are used to protect active magnetic bearings in the event of rotor delevitation. To gain insight into the design, selection and operation of BBs, a model that can accurately simulate the behaviour of a rotor and BB system during a rotor delevitation event is required. This paper presents a model that can accurately simulate rotor delevitation and predict the occurrence of forward and backward whirl. The model, called BBSim, is based on the integration of a number of simpler models for each of the important mechanisms that govern the behaviour of a BB system. Estimation of parameters that are critical in determining the rotor’s behaviour inside the BB clearance is achieved using empirical methods. Rotor orbit plots confirm the model’s capability of simulating the characteristic motions of rotor delevitation.

The dynamic reconfiguration of an FPGA has many advantages, but the overhead from the process reduces the functional density of applications. Functional density is an indication of the composite benefits a reconfigured application obtains above its generic counterpart and measures the computational throughput per unit hardware resources. Typically, only quasi-static applications obtain a functional density advantage by dynamically reconfiguring its parameters. Contributing to the functional density reduction of applications with tight time constraints is the overhead to generate a new configuration, and the time it takes to load it onto the device. Normally these applications have to reuse their hardware numerous times between configurations before obtaining a functional density advantage. The most promising reconfiguration method to improve functional density with minimal hardware reuse was one that extracts certain characteristics from the bitstream and then implements a bitstream specialiser that generates new hardware at bit-level while the device is being reconfigured. While it was shown that this method allows reconfiguration of an application in real-time, its effect on functional density was not determined. This paper will show that a significant increase in functional density can be achieved for applications where reconfiguration is required before the next execution cycle of the application.CATEGORIES: · Hardware ~ Reconfigurable logic and FPGAs · Computer systems organization ~ Other architectures

Despite the many advantages run-time reconfiguration of FPGAs brings to the table, its usage is mostly limited to quasi-static applications. This is either due to the throughput of the reconfiguration process, or the time required to create new hardware. In order to optimise the former, the literature proposes a block RAM (BRAM)-based architecture in which a new configuration is stored in localised memory and reconfiguration is facilitated by a controller implemented in the FPGA fabric. The limitation of this architecture is that only a subset of configurations can be stored. When new hardware is required, the slow synthesis process (or a part thereof) has to be repeated for each new configuration. Various third-party tools aim to mitigate this overhead, but since the bitstream is shrouded in obscurity, all rely on a layer of abstraction that make them unusable in real-time. To address this issue, this paper presents a novel method to parse and analyse a Xilinx® FPGA bitstream to extract certain characteristics. It is shown how these characteristics could be used to design and implement a bitstream specialiser, capable of taking a bitstream and modifying the configuration bits of lookup tables in real-time. CATEGORIES: · Hardware ~ Reconfigurable logic and FPGAs · Computer systems organization ~ Other architectures

MEng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus Decomposing a complex shape into visually significant parts comes naturally for humans, and turns out to be very useful in areas such as shape analysis, shape matching, recognition, topology extraction, collision detection and other geometric processing methods [1]. After analysis it was found that the Minimum Near-Convex Decomposition (MNCD) method [2] is one of the most promising algorithms currently available that shows room for improvement. The focus of this dissertation is to make an improvement on the time it takes to decompose a complex shape, while keeping the decomposition (number of parts) results the same. One improvement that was implemented was to neglect the Morse function, as this takes a long time to execute. Another improvement was to make use of Delaunay Triangulation (DT) instead of considering all of the vertices, as no overlapping will take place and the need for the non-overlapping matrix is no longer necessary. Experimental results show that an average time reduction of 58%, but an increase in the number of parts. Thus there is an improvement made on the duration of the algorithm, but there is room to improve on the total amount of parts obtained after decomposition. Masters

MEng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus Feature selection is crucial for increasing the performance of predictive models in both classification accuracy and model training time. For high-dimensional data, feature selection becomes ever more necessary to select adequate features which complement the predictive model of choice. Filter, wrapper and embedded feature selection techniques are among the most popular algorithms to solve the feature selection conundrum of irrelevant and redundant features reducing the performance of classification models. It is also common to find hybrid techniques which combines filter, wrapper and embedded techniques to construct more robust feature selection algorithms. This study is dedicated to reveal the ongoing improvement in the field of feature selection and to dissect six different feature selection algorithms for a detailed insight into their success for high dimensional data, specifically gene expression microarrays. The six algorithms for this study are: (i) three filter methods mRMR (min-Redundancy Max-Relevance), FCBF# (Fast Correlation Based Filter) and ORFS (Orthogonal Relevance Feature Selection), (ii) two wrapper methods FRBPSO (Fuzzy Rule Based Particle Swarm Optimisation) and SVM-RFE (Support Vector Machine-Recursive Feature Elimination), and (iii) one embedded method SBMLR (Sparse Multinomial Logistic Regression via Bayesian L1 regularisation). The three filter methods are adapted into suitable hybrid techniques and multiple associative measures are explored to determine the best performance per algorithm. All algorithms include the pre-processing techniques MDL discretisation and SIS to explore their improvements and shortcomings. The performance per algorithm is based on their ability to improve classification accuracy with the least amount of features possible and compared to one another. After comparison, the algorithms best suited for classification improvement, computation speed advantage and feature removal capability are revealed. Thereafter, a case study involving plant foliage features where the amount of features greatly outnumber the number of samples, denoted by p >> n, is used to compliment the findings. The use of pre-processing techniques proved to be and SVM-RFE (Support Vector Machine-Recursive Feature Elimination), and (iii) one embedded method SBMLR (Sparse Multinomial Logistic Regression via Bayesian L1 regularisation). The three filter methods are adapted into suitable hybrid techniques and multiple associative measures are explored to determine the best performance per algorithm. All algorithms include the pre-processing techniques MDL discretisation and SIS to explore their improvements and shortcomings. The performance per algorithm is based on their ability to improve classification accuracy with the least amount of features possible and compared to one another. After comparison, the algorithms best suited for classification improvement, computation speed advantage and feature removal capability are revealed. Thereafter, a case study involving plant foliage features where the amount of features greatly outnumber the number of samples, denoted by p >> n, is used to compliment the findings. The use of pre-processing techniques proved to be crucial regarding improved classification accuracy and reduced computation time. Out of all six algorithms, mRMR, and SVM-RFE proved the most promising. Masters

PhD (Computer and Electronic Engineering), North-West University, Potchefstroom Campus, 2019. This thesis presents the cooperative control of an active magnetic bearing (AMB) and sensorless drive system. Traditionally, these two systems are considered as independent modular components within a larger system. The control of both systems requires angular speed information. Principally, the speed is required in the unbalance control of the AMB and the speed control of the drive. Furthermore, the speed is also required in both systems to compute cross-coupling terms for feed forward cancellation. Historically, the speed is estimable in both systems, but only above a threshold speed. The speed estimation in the AMB relied on the orbit generated due to unbalance. The speed estimation of the sensorless drive (of a surface mounted PMSM) relied upon the generated back-emf. This resulted in the sensorless drive system requiring an open-loop start-up procedure. It has been found by utilizing a novel method which relies on the disturbance force on the rotor, due to unbalanced magnetic pull, that it is possible for the AMB system to estimate the angular position of the rotor from standstill up to an upper limit in speed. Sharing this estimated angular position with the drive system, it is possible to skip the open-loop start-up of the drive and start it in vector controlled mode. The drive is switched to sensorless drive mode before the angular position accuracy upper limit is reached. After this switch over, the estimated angle by the sensorless vector control is in turn shared with the AMB system for unbalance control. The sensorless drive control algorithm has been reorganized to integrate in such a manner with the AMB’s estimated angular position so that a smooth bump less transfer results during the switch over. The main contributions of this work are the disturbance force model identification by the AMB system for estimation of the angular position and the supervisory coordination of the two controlled systems to act cooperatively. The key structure proposed for consolidation of the shared state information between the two systems is by integration into a phase-locked loop (PLL) National Research Foundation (NRF) Doctoral

Physics-based battery models can be very complex and require careful experimental validation, but yield greater insights into the internal processes and their interactions than other cell modelling approaches. The complexity is associated with the large number of input parameters that (a) have varying degrees of identifiability, (b) can be constant or varying, and (c) appear in complex combinations with the model variables and each other. The current work studies this complexity by proposing a unique stepwise approach and purposefully addresses the computational cost associated with estimating these parameters. An initial set of 50 input parameters is reduced to only 8 highly influential parameters that can be subjected to parameter optimization. Elementary effects analysis using the Morris method is applied and demonstrates that the electrode kinetic parameters dominate the cell’s voltage response as simulated by the model. These influential parameters are subjected to variance-based sensitivity analysis using Monte Carlo experiments and Jansen’s formulae for variance decomposition. The first-order and total sensitivities during various modes of operation indicate that the charge transfer coefficients and effective exchange current densities have the most influence on the model error and should be subjected to further optimization. The model error’s sensitivity also reveals high parameter identifiability within subsets of the experimental data and indicates that some parameter values might be valid for longer timescales rather than shorter timescales. The proposed stepwise approach can be applied to any complex physics-based cell model regardless of the cell’s chemistry, format or form factor.

This paper investigates the effectiveness of current test methods and procedures associated with insulation testing on the busbar of Air Insulated Switchgear (AIS) during scheduled planned maintenance. An improved test method proposes combining traditional AC voltage withstand test with Partial Discharge (PD) measurement. Three PD sensors were considered, namely Airborne Acoustic (AA), Transient Earth Voltage (TEV) and High Frequency Current Transformer (HFCT). The results confirm that HFCT is most sensitive and revealed significant difference at higher voltages between the six AIS samples. The TEV results produced consistent readings with smaller differences at higher voltages. The Ultra-sound instrument was found to be extremely sensitive but lack the capability to produce reliable measurement readings. Peak TEV measurements indicate a noticeable difference between older and newer samples

Heat is one of the major causes of degradation which can cause a reduction in performance, reliability and life span of miniature circuit breakers (MCBs) [1]. In order to diagnose the condition of electrical equipment, the thermal stress of that equipment should be known [2]. Due to the compact design of MCB’s it is important to minimize or dissipate the heat that builds up inside MCBs during normal use. With time electrical components and contact surfaces heats up and begin to deteriorate, this increase in heat is often a product of increased electrical resistance which is a result of degradation [3]. Infrared thermography is a condition monitoring technique that can be applied to electrical distribution boards but because of design limitations MCBs are stacked next to each other on a rail in a panel allowing a direct view only from the front. Is the surface temperature of the MCB a reflection of the internal degradation or are there other factors that need to be considered to accurately diagnose the MCB? The following methodology for the exploratory study was followed to determine the heat profile after degradation, a 20A hydraulic-magnetic MCB was switched 9000 times at rated voltage and current. The volt drop across the conduction path as well as thermal images of the front and sides were taken after every 500 switching operations, after the MCB was allowed to heat up for 1 hour at rated current. When comparing the surface temperature rise above ambient of the three sides the results shows that the number of operations does not influence the surface temperature of the front and the left (hydraulic magnetic unit) as much as the right (conduction path). The aim of this paper is to use a lumped model and compare with empirical infrared thermographic results relating to the condition of MCBs and to: Define the MCB surface heat distribution after a number of switching operations at rated current. Determine whether the frontal surface temperature of the MCB is an accurate indicator of internal degradation. Determine a field methodology to determine degradation and to anticipate failure of MCBs? Due to the construction of the MCB the front, left and right sides give different heat patterns that complicate matter with regards to determining the correct internal health state of the MCB

A number of gearbox failures can be attributed to lubricant related problems. One measure of the condition of gearbox oil is its viscosity. In electrically powered systems, motor current signal analysis allows online estimation of the viscosity of gearbox oil without requiring additional sensors. Previous work on this problem entailed monitoring the power (and change in power) of sidebands of the shaft frequency in the induction motor current spectrum. Sideband frequencies in the current spectrum can however be influenced by other potential problems in the electromechanical system ranging from bearing faults to gearbox teeth damage.\ud \ud \ud Changes in the lubricant viscosity result in changes in the mechanical and thermal losses in the system. These small deviations in the mechanical and thermal losses in the system become visible in the ratio of the electrical energy demanded by the induction motor to the kinetic energy of the rotating mechanical parts. Speed and load invariance can be ensured by normalizing the measured energy ratio with lookup table values obtained when the system attained thermal equilibrium. Speed or load perturbations in the system give rise to small deviations in the normalized energy ratio curve. The distributions of these deviations are significantly different (in a statistical sense) for different oil viscosity values.

For a humanitarian landmine detection system to be of practical value it has to meet the following criteria: detection must be performed in real-time from a moving platform; plastic mines must be reliably detected; and the false alarm rate should be kept to a minimum. Ground penetrating radar is a promising technology for the detection of landmines with low metal content. The improved algorithms presented in this paper can meet the above mentioned criteria to a large extent. Clutter removal is performed by an adaptive filter which removes virtually all background clutter in real time and can adapt to changing soil characteristics. Classification is performed by a new rule-based classifier capable of detecting both metal and plastic anti-tank mines. Despite the fact that the algorithms are currently implemented in Matlabr, mine detection can be performed on a vehicle moving at 9 km/h http://www.saiee.org.za/displaycustomlink.aspx?name=AfricaResearchJournal

Kabelo, I'm not sure if this journal is open access??? The presence of landmines poses a worldwide humanitarian problem. Often, these mines are difficult to detect with metal detectors. Ground penetrating radar (GPR) is a promising technology for the detection of landmines with low metal content. Automatic landmine detection typically consists of two steps, namely preprocessing (or clutter removal) and classification. In this paper the clutter removal algorithm consists of a nonlinear frequency domain filter followed by principal component based filtering. Principal component analysis is performed in the frequency domain to build a background model for the clutter. The latter model is removed from the observed data in the log-frequency domain in order to preserve the phase component of the spectrum. Finally, the data is normalized and transformed to the time domain. The results presented in this paper show a marked improvement in the ability to remove general background clutter. Classification is performed on the basis of the prediction performance of neural network time-series models of the various classes of GPR responses. The classification system can correctly identify the position of metal anti-tank (AT) mines. It can also recognize specific examples of low metal AT and (anti-personnel) AP mines, but does have a low generalization ability for such mines http://www.saiee.org.za/displaycustomlink.aspx?name=AfricaResearchJournal

The stability margin of a two degree-of-freedom self-sensing active magnetic bearing (AMB) is estimated by means of μ -analysis. The specific self-sensing algorithm implemented in this study is the direct current measurement method. Detailed black-box models are developed for the main subsystems in the AMB by means of discrete-time system identification. In order to obtain models for dynamic uncertainty in the various subsystems in the AMB, the identified models are combined to form a closed-loop model for the self-sensing AMB. The response of this closed-loop model is then compared to the original AMB’s response and models for the dynamic uncertainty are empirically deduced. Finally, the system’s stability margin for the modelled uncertainty is estimated by means of μ -analysis. The results show that μ -analysis is ill-equipped to estimate the stability margin of a nonlinear system

Coal ore isn't a uniform material. In order to optimize the coal liberation process it is necessary to classify a coal ore sample into its constituent components as quickly and cheaply possible. This paper investigates whether it is feasible to employ image processing and pattern recognition to segment a photographic image of coal ore into its various mineral components prior to the sample being crushed. The key to solving this classification problem is to model the visual texture of the various coal components by means of a low-dimensional texture space consisting of two main dimensions, namely: roughness and regularity. The regularity of each texture is estimated by means of a novel model-based approach. The distribution of the various coal components in the resultant feature space is modelled by means of a mixtures model and a simple nearest-neighbour decision rule is used to classify each pixel in the image. The performance of the classification system is encouraging and shows the feasibility of our idea.

A fair amount of research has been done on the modeling of the oculomotor system. Models of the oculomotor neural circuitry that include the effect of alcohol are less plentiful in literature. We investigated the correlation between blood-alcohol levels and the response characteristics of the optical-oculomotor system. Video recordings of eye movements were regarded as the main sensor, with a breathalyzer for verification. System identification was done based on the transfer function of a proposed delay-based model. The correlation of coefficients of the model to increasing blood-alcohol concentration was seen to display behavior that was unique to each test.

Reconfigurable computing is a paradigm in computing architecture that refers to the practice of using interchangeable hardware modules to enhance the performance of conventional Von-Neumann style computing. Despite the numerous advantages of reconfiguration, it is only suitable for quasi-static applications with slowly changing reconfiguration criteria. In general, it is only advantageous to reconfigure if the execution time exceeds recon guration time. Since the execution time of real-time systems are quite limited, the control of dynamic non-linear systems are typically not reconfigured. Instead, adaptivity is mostly gained from reading coefficients or gains from memory. Where different controller architectures are required, the route of parallel implementation could be taken, switching between architectures. The drawback of this approach is an increase in area required to implement the controllers. Reconfiguration on the other hand could allow the different control architectures to be swapped on the y without interrupting operation of the controller, while minimizing the area required. Unfortunately, this process is limited by the overhead introduced by the reconfiguration. Even though various survey papers exist on the topic of reconfiguration, none really focus on methods to reduce the cost of reconfiguration. This survey summarizes different means of reducing configuration overhead in an attempt to allow reconfiguration of applications with limited execution time. A block RAM- based (BRAM) architecture is proposed as the optimal architecture for reconfiguring dynamic applications. As an example, this architecture is used to discuss the methodology used to design a reconfigurable PID controller Technology and Human Re- sources for Industry Programme (THRIP) and Oppenheimer Memo- rial Trust Grant (Ref. 19328/01) http://www.ifac-papersonline.net/

This paper presents the cooperative control between the active magnetic bearing system and the permanent magnet synchronous drive. More specifically, the coupling between the vibration control and sensorless drive control is considered since both systems have the ability to estimate the angular speed. The coupling of the estimated speed in each controller is considered. Also, a modified start-up procedure for the sensorless vector control had to be developed when the drive is combined with the AMB system to accommodate the low friction which does not yield damping of the speed perturbation at start-up. Simulation results are presented for the developed control using detailed models for both systems http://www.ifac-papersonline.net/

Energy is a universal concept that can be used across physical domains to describe complex large-scale industrial systems. This brief survey and framework gives a perspective on energy as a unifying domain for system modelling, supervision, and control. Traditionally, modelling and control problems have been approached by adopting a signal-processing paradigm. However, this approach becomes problematic when considering non-linear systems. A behavioural viewpoint, which incorporates energy as basis for modelling and control, is considered a viable solution. Since energy is seen as a unifying concept, its relationship to Euler-Lagrange equations, state space representation, and Lyapunov functions is discussed. The connection between control and process supervision using passivity theory coupled with a system energy balance is also established. To show that complex industrial systems comprising multiple energy domains can be modelled by means of a single electric circuit, its application to a large-scale thermo-hydraulic system is presented. Next, a simple non-linear transmission impedance electric circuit is used to illustrate how energy can be used to not only describe a system, but also serve as basis for system optimisation. An energy-based framework is proposed whereby energy is used as a unifying domain to work in, to analyse, and to optimise large-scale industrial systems http://www.ifac-papersonline.net/

The relatively low cost of distribution transformers dictates that the cost of any condition monitoring system for such transformers should be as low as possible. Premature failure of any transformer in an electrical grid does invariably lead to customer inconvenience and costs. Partial discharges are a key indicator of the state of health of a transformer's insulation and can be non-intrusively monitored by means of transient earth voltage (TEV) sensors, thereby minimizing installation costs of the condition monitoring system. The high bandwidth of partial discharge pulses traditionally requires very high speed (and very expensive) digitizers. Sub-Nyquist sampling can be employed to reduce the required sampling rate to monitor insulation degradation. In this paper, partial discharge pulses measured with TEV sensors with a bandwidth of 40 MHz are sampled at 12.8 kHz with the aim of constructing phaseresolved partial discharge (PRPD) patterns as cheaply as possible. The resulting PRPD patterns are then sent to a support vector machine based classifier to detect various insulation defects timeously. Simulation results show that this approach is able to detect different types of partial discharge events

In a renewable energy system, lead-acid batteries are required to last as long as possible whilst providing deep discharges and are usually charged in an intermittent manner. This type of operation is severely detrimental to the electrodes and optimal performance cannot be guaranteed. Correct operation can be achieved if a charge controller receives accurate information about the state of the battery and its cells. Accurate approximation of the available active surface area is essential to optimal performance since it determines the overall capacity and reaction rates of various secondary processes. This paper improves on the current understanding of the active surface area in a lead-acid cell during discharge by comparing three approximation methods. Two methods use the state-of-charge based on porosity whilst the other method uses the charge per unit volume to determine the remaining capacity. These three methods are used in a comprehensive macro-homogeneous electrochemical model which accounts for initial temperature differences. Reported experimental behaviour is used to evaluate each method and recommendations are made regarding the selection of associated parameter values. The approximation method by Cugnet et al. (2009) is the most suitable because it always results in a concave morphology. The current work can aid in the development of a health-conscious battery management system HySA Infrastructure http://www.ifac-papersonline.net/

Good assessment is an essential component of good learning. Marking test and exam papers for large classes is however a time-consuming and repetitive task. This paper reports on the design and implementation of a computer based assessment (CBA) system suitable for undergraduate engineering modules. This prototype CBA system is capable of automatically grading relatively open-ended analysis and design problems and is also relatively user friendly. This CBA system consists of two separate subsystems, namely a test delivery subsystem and an automatic grading subsystem (which is the contribution of this paper). Automatic grading of the student answer papers is performed in Matlab . The grading subsystem can automatically grade typed student answers containing symbolic mathematical expressions as well as numeric answers. Grading is performed accurately, consistently and very fast compared to a human lecturer. Students are given partial credit for partially correct answers and feedback is given in a very intuitive form quite close to the traditional manner in a pen-and-paper test. The system has been successfully used to take final exams in a second year engineering module on basic circuit analysis http://www.saiee.org.za/displaycustomlink.aspx?name=AfricaResearchJournal http://www.saiee.org.za/displaycustomlink.aspx?name=2013:%20Vol%20104%20No%204#4

Despite the numerous advantages reconfigurable computing adds to a system, it is only advantageous if the execution time exceeds the configuration time. As a result of the long configuration time, reconfiguration is only suitable for quasi-static applications. Due to the additional overhead required for communication, the bus-based architectures most commonly used to connect the configuration controller to the memory contribute to the configuration time. A method proposed to ameliorate this overhead is an architecture utilizing localized block RAM (BRAM), directly connected to the configuration controller to store the configuration data. The drawback of this method is that the BRAM is extremely limited and only a discrete set of configurations can be stored. This paper is a work in progress and proposes a hardware reconfiguration architecture that addresses the size limitation of the localized BRAM-architecture by using parameterizable configuration. This will allow a single bitstream stored in the BRAM to be specialized according to certain parameters, which will be used to reconfigure the device. This will migrate reconfigurable computing to more dynamic applications. The architecture proposed in this paper will be validated using real-time PID control of a five degree of freedom active magnetic bearing system.

A state-space model with uncertainties for an active magnetic bearing energy storage flywheel system (Fly-UPS) is developed. A multivariable robust H∞ controller for the Fly-UPS is then synthesised. Different weighting schemes are explained and used for H∞ controller synthesis. Furthermore, the additive uncertainties between the nominal simulation model and the physical model at varied rotational speeds are characterised. Stability robustness for additive uncertainties is verified via the gain/phase margin stability robustness criterion. The performance robustness (disturbance attenuation) is analysed by assessing model sensitivity with the ISO/CD 14839-3 sensitivity standard.