Active inverter output filtering methods

Frequency converters are widely used in the industry to enable better controllability and efficiency of variable speed AC motor drives. Despite these advantages, certain challenges concerning the inverter and motor interfacing have been present for decades. As insulated gate bipolar transistors entered the market, the inverter output voltage transition rate significantly increased compared with their predecessors. Inverters operate based on pulse width modulation of the output voltage, and the steep voltage edge fed by the inverter produces a motor terminal overvoltage. The overvoltage causes extra stress to the motor insulation, which may lead to a prematuremotor failure. The overvoltage is not generated by the inverter alone, but also by the sum effect of the motor cable length and the impedance mismatch between the cable and the motor. Many solutions have been shown to limit the overvoltage, and the mainstream products focus on passive filters. This doctoral thesis studies an alternative methodology for motor overvoltage reduction. The focus is on minimization of the passive filter dimensions, physical and electrical, or better yet, on operation without any filter. This is achieved by additional inverter control and modulation. The studied methods are implemented on different inverter topologies, varying in nominal voltage and current.For two-level inverters, the studied method is termed active du/dt. It consists of a small output LC filter, which is controlled by an independent modulator. The overvoltage is limited by a reduced voltage transition rate. For multilevel inverters, an overvoltage mitigation method operating without a passive filter, called edge modulation, is implemented. The method uses the capability of the inverter to produce two switching operations in the same direction to cancel the oscillating voltages of opposite phases. For parallel inverters, two methods are studied. They are both intended for two-level inverters, but the first uses individual motor cables from each inverter while the other topology applies output inductors. The overvoltage is reduced by interleaving the switching operations to produce a similar oscillation accumulation as with the edge modulation. The implementation of these methods is discussed in detail, and the necessary modifications to the control system of the inverter are presented. Each method is experimentally verified by operating industrial frequency converters with the modified control. All the methods are found feasible, and they provide sufficient overvoltage protection. The limitations and challenges brought about by the methods are discussed.

Power semiconductor nonlinearities in active du/dt output filtering

This doctoral thesis introduces an improved control principle for active du/dt output filtering in variable-speed AC drives, together with performance comparisons with previous filtering methods. The effects of power semiconductor nonlinearities on the output filtering performance are investigated. The nonlinearities include the timing deviation and the voltage pulse waveform distortion in the variable-speed AC drive output bridge. Active du/dt output filtering (ADUDT) is a method to mitigate motor overvoltages in variable-speed AC drives with long motor cables. It is a quite recent addition to the du/dt reduction methods available. This thesis improves on the existing control method for the filter, and concentrates on the lowvoltage (below 1 kV AC) two-level voltage-source inverter implementation of the method. The ADUDT uses narrow voltage pulses having a duration in the order of a microsecond from an IGBT (insulated gate bipolar transistor) inverter to control the output voltage of a tuned LC filter circuit. The filter output voltage has thus increased slope transition times at the rising and falling edges, with an opportunity of no overshoot. The effect of the longer slope transition times is a reduction in the du/dt of the voltage fed to the motor cable. Lower du/dt values result in a reduction in the overvoltage effects on the motor terminals. Compared with traditional output filtering methods to accomplish this task, the active du/dt filtering provides lower inductance values and a smaller physical size of the filter itself. The filter circuit weight can also be reduced. However, the power semiconductor nonlinearities skew the filter control pulse pattern, resulting in control deviation. This deviation introduces unwanted overshoot and resonance in the filter. The controlmethod proposed in this thesis is able to directly compensate for the dead time-induced zero-current clamping (ZCC) effect in the pulse pattern. It gives more flexibility to the pattern structure, which could help in the timing deviation compensation design. Previous studies have shown that when a motor load current flows in the filter circuit and the inverter, the phase leg blanking times distort the voltage pulse sequence fed to the filter input. These blanking times are caused by excessively large dead time values between the IGBT control pulses. Moreover, the various switching timing distortions, present in realworld electronics when operating with a microsecond timescale, bring additional skew to the control. Left uncompensated, this results in distortion of the filter input voltage and a filter self-induced overvoltage in the form of an overshoot. This overshoot adds to the voltage appearing at the motor terminals, thus increasing the transient voltage amplitude at the motor. This doctoral thesis investigates the magnitude of such timing deviation effects. If the motor load current is left uncompensated in the control, the filter output voltage can overshoot up to double the input voltage amplitude. IGBT nonlinearities were observed to cause a smaller overshoot, in the order of 30%. This thesis introduces an improved ADUDT control method that is able to compensate for phase leg blanking times, giving flexibility to the pulse pattern structure and dead times. The control method is still sensitive to timing deviations, and their effect is investigated. A simple approach of using a fixed delay compensation value was tried in the test setup measurements. The ADUDT method with the new control algorithm was found to work in an actual motor drive application. Judging by the simulation results, with the delay compensation, the method should ultimately enable an output voltage performance and a du/dt reduction that are free from residual overshoot effects. The proposed control algorithm is not strictly required for successful ADUDT operation: It is possible to precalculate the pulse patterns by iteration and then for instance store them into a look-up table inside the control electronics. Rather, the newly developed control method is a mathematical tool for solving the ADUDT control pulses. It does not contain the timing deviation compensation (from the logic-level command to the phase leg output voltage), and as such is not able to remove the timing deviation effects that cause error and overshoot in the filter. When the timing deviation compensation has to be tuned-in in the control pattern, the precalculated iteration method could prove simpler and equally good (or even better) compared with the mathematical solution with a separate timing compensation module. One of the key findings in this thesis is the conclusion that the correctness of the pulse pattern structure, in the sense of ZCC and predicted pulse timings, cannot be separated from the timing deviations. The usefulness of the correctly calculated pattern is reduced by the voltage edge timing errors. The doctoral thesis provides an introductory background chapter on variable-speed AC drives and the problem of motor overvoltages and takes a look at traditional solutions for overvoltage mitigation. Previous results related to the active du/dt filtering are discussed. The basic operation principle and design of the filter have been studied previously. The effect of load current in the filter and the basic idea of compensation have been presented in the past. However, there was no direct way of including the dead time in the control (except for solving the pulse pattern manually by iteration), and the magnitude of nonlinearity effects had not been investigated. The enhanced control principle with the dead time handling capability and a case study of the test setup timing deviations are the main contributions of this doctoral thesis. The simulation and experimental setup results show that the proposed control method can be used in an actual drive. Loss measurements and a comparison of active du/dt output filtering with traditional output filtering methods are also presented in the work. Two different ADUDT filter designs are included, with ferrite core and air core inductors. Other filters included in the tests were a passive du/dtfilter and a passive sine filter. The loss measurements incorporated a silicon carbide diode-equipped IGBT module, and the results show lower losses with these new device technologies. The new control principle was measured in a 43 A load current motor drive system and was able to bring the filter output peak voltage from 980 V (the previous control principle) down to 680 V in a 540 V average DC link voltage variable-speed drive. A 200 m motor cable was used, and the filter losses for the active du/dt methods were 111W–126 W versus 184 W for the passive du/dt. In terms of inverter and filter losses, the active du/dt filtering method had a 1.82-fold increase in losses compared with an all-passive traditional du/dt output filter. The filter mass with the active du/dt method was 17% (2.4 kg, air-core inductors) compared with 14 kg of the passive du/dt method filter. Silicon carbide freewheeling diodes were found to reduce the inverter losses in the active du/dt filtering by 18% compared with the same IGBT module with silicon diodes. For a 200 m cable length, the average peak voltage at the motor terminals was 1050 V with no filter, 960 V for the all-passive du/dt filter, and 700 V for the active du/dt filtering applying the new control principle.

Metabolism of diglycine and triglycine by non-filtering kidneys

We have studied the metabolism of diglycine and triglycine in the isolated non-filtering rat kidney. Kidneys from adult male Wistar Kyoto rats weighing 250-350 g were perfused with Krebs-Henseleit solution containing either 1 mM diglycine or triglycine. The analysis of the peptide residues and their components was performed using an amino acid microanalyzer utilizing ion exchange chromatography. Diglycine was degraded to a final concentration of 0.09 mM after 120 min (91%); this degradation occurred predominantly during the first hour, with a 56% reduction of the initial concentration. The metabolism of triglycine occurred similarly, with a final concentration of 0.18 mM (82%); during the first hour there was a 67% reduction of the initial concentration of the tripeptide. Both peptides produced glycine in increasing concentrations, but there was a slightly lower recovery of glycine, suggesting its utilization by the kidney as fuel. The hydrolysis of triglycine also produced diglycine, which was also hydrolyzed to glycine. The results of the present study show the existence of functional endothelial or contraluminal membrane peptidases which may be important during parenteral nutrition.

Adaptive Markov chain Monte Carlo and Bayesian filtering for state space models

This thesis is concerned with the state and parameter estimation in state space models. The estimation of states and parameters is an important task when mathematical modeling is applied to many different application areas such as the global positioning systems, target tracking, navigation, brain imaging, spread of infectious diseases, biological processes, telecommunications, audio signal processing, stochastic optimal control, machine learning, and physical systems. In Bayesian settings, the estimation of states or parameters amounts to computation of the posterior probability density function. Except for a very restricted number of models, it is impossible to compute this density function in a closed form. Hence, we need approximation methods. A state estimation problem involves estimating the states (latent variables) that are not directly observed in the output of the system. In this thesis, we use the Kalman filter, extended Kalman filter, Gauss–Hermite filters, and particle filters to estimate the states based on available measurements. Among these filters, particle filters are numerical methods for approximating the filtering distributions of non-linear non-Gaussian state space models via Monte Carlo. The performance of a particle filter heavily depends on the chosen importance distribution. For instance, inappropriate choice of the importance distribution can lead to the failure of convergence of the particle filter algorithm. In this thesis, we analyze the theoretical Lᵖ particle filter convergence with general importance distributions, where p ≥2 is an integer. A parameter estimation problem is considered with inferring the model parameters from measurements. For high-dimensional complex models, estimation of parameters can be done by Markov chain Monte Carlo (MCMC) methods. In its operation, the MCMC method requires the unnormalized posterior distribution of the parameters and a proposal distribution. In this thesis, we show how the posterior density function of the parameters of a state space model can be computed by filtering based methods, where the states are integrated out. This type of computation is then applied to estimate parameters of stochastic differential equations. Furthermore, we compute the partial derivatives of the log-posterior density function and use the hybrid Monte Carlo and scaled conjugate gradient methods to infer the parameters of stochastic differential equations. The computational efficiency of MCMC methods is highly depend on the chosen proposal distribution. A commonly used proposal distribution is Gaussian. In this kind of proposal, the covariance matrix must be well tuned. To tune it, adaptive MCMC methods can be used. In this thesis, we propose a new way of updating the covariance matrix using the variational Bayesian adaptive Kalman filter algorithm.

Clinical assessment of the effect of digital filtering on the detection of ventricular late potentials

Ventricular late potentials are low-amplitude signals originating from damaged myocardium and detected on the body surface by ECG filtering and averaging. Digital filters present in commercial equipment may interfere with the ability of arrhythmia stratification. We compared 40-Hz BiSpec (BI) and classical 40- to 250-Hz band-pass Butterworth bidirectional (BD) filters in terms of impact on time domain variables and diagnostic properties. In a transverse retrospective age-adjusted case-control study, 221 subjects with sinus rhythm without bundle branch block were divided into three groups after signal-averaged ECG acquisition: GI (N = 40), clinically normal controls, GII (N = 158), subjects with coronary heart disease without sustained monomorphic ventricular tachycardia (SMVT), and GIII (N = 23), subjects with heart disease and documented SMVT. Conventional variables analyzed from vector magnitude data after averaging to 0.3 µV final noise were obtained by application of each filter to the averaged signal, and evaluated in pairs by numerical comparison and by diagnostic agreement assessment, using conventional and optimized thresholds of normality. Significant differences were found between BI and BD variables in all groups, with diagnostic results showing significant disagreement between both filters [kappa value of 0.61 (P<0.05) for GII and 0.31 for GIII (P = NS)]. Sensitivity for SMVT was lower with BI than with BD (65.2 vs 91.3%, respectively, P<0.05). Filters provided significantly different numerical and diagnostic results and the BI filter showed only limited clinical application to risk stratification of ventricular arrhythmia.

Physicochemical and sensory changes in aged sugarcane spirit submitted to filtering with activated carbon filter

Sugarcane spirit is a drink considered as a national symbol of Brazil. It is produced by large producers and by about 30 thousand small and medium home-distilling producers dispersed throughout the country. The copper originating from the home-distillers can become a serious problem since at high concentrations in beverages it may cause serious human health problems. Therefore, the objective of this study was to investigate the influence of the activated carbon used in commercial filters on the physicochemical and sensory characteristics of aged sugarcane spirit. Analyses of copper, dry extract, alcoholic degree, higher alcohols, volatile acids, aldehydes, esters, furfural, and methanol were performed. The sensory evaluation was performed by seven selected trained judges, who analyzed the yellow color, woody aroma and flavor, and intensity of alcoholic aroma and flavor of the cane spirit before and after the filtration process. The sensory tests were carried out using a 9 cm non-structured intensity scale. A reduction was observed in all compounds analyzed physicochemically, except for the esters, which increased after filtration. This increase is probably due to the esterification of the alcohols and acids present. According to the sensory results obtained, a reduction was observed in the intensity of the yellow color, aroma, and wood flavor characteristics, the major characteristics of the aging process.

Variational Ensemble Kalman Filtering in Hydrology

The current thesis manuscript studies the suitability of a recent data assimilation method, the Variational Ensemble Kalman Filter (VEnKF), to real-life fluid dynamic problems in hydrology. VEnKF combines a variational formulation of the data assimilation problem based on minimizing an energy functional with an Ensemble Kalman filter approximation to the Hessian matrix that also serves as an approximation to the inverse of the error covariance matrix. One of the significant features of VEnKF is the very frequent re-sampling of the ensemble: resampling is done at every observation step. This unusual feature is further exacerbated by observation interpolation that is seen beneficial for numerical stability. In this case the ensemble is resampled every time step of the numerical model. VEnKF is implemented in several configurations to data from a real laboratory-scale dam break problem modelled with the shallow water equations. It is also tried in a two-layer Quasi- Geostrophic atmospheric flow problem. In both cases VEnKF proves to be an efficient and accurate data assimilation method that renders the analysis more realistic than the numerical model alone. It also proves to be robust against filter instability by its adaptive nature.

Rigid body dynamics and Kalman filtering

X-ray computed log tomography has always been applied for qualitative reconstructions. In most cases, a series of consecutive slices of the timber are scanned to estimate the 3D image reconstruction of the entire log. However, the unexpected movement of the timber under study influences the quality of image reconstruction since the position and orientation of some scanned slices can be incorrectly estimated. In addition, the reconstruction time remains a significant challenge for practical applications. The present study investigates the possibility to employ modern physics engines for the problem of estimating the position of a moving rigid body and its scanned slices which are subject to X-ray computed tomography. The current work includes implementations of the extended Kalman filter and an algebraic reconstruction method for fan-bean computer tomography. In addition, modern techniques such as NVidia PhysX and CUDA are used in current study. As the result, it is numerically shown that it is possible to apply the extended Kalman filter together with a real-time physics engine, known as PhysX, in order to determine the position of a moving object. It is shown that the position of the rigid body can be determined based only on reconstructions of its slices. However, the simulation of the body movement sometimes is subject to an error during Kalman filter employment as PhysX is not always able to continue simulating the movement properly because of incorrect state estimation.

Histogram filtering as a tool in variational Monte Carlo optimization

Optimization of wave functions in quantum Monte Carlo is a difficult task because the statistical uncertainty inherent to the technique makes the absolute determination of the global minimum difficult. To optimize these wave functions we generate a large number of possible minima using many independently generated Monte Carlo ensembles and perform a conjugate gradient optimization. Then we construct histograms of the resulting nominally optimal parameter sets and "filter" them to identify which parameter sets "go together" to generate a local minimum. We follow with correlated-sampling verification runs to find the global minimum. We illustrate this technique for variance and variational energy optimization for a variety of wave functions for small systellls. For such optimized wave functions we calculate the variational energy and variance as well as various non-differential properties. The optimizations are either on par with or superior to determinations in the literature. Furthermore, we show that this technique is sufficiently robust that for molecules one may determine the optimal geometry at tIle same time as one optimizes the variational energy.

A computational fluid dynamic study on the filtering mechanics in suspension feeding marine invertebrates

Les suspensivores ont la tâche importante de séparer les particules de l'eau. Bien qu'une grande gamme de morphologies existe pour les structures d'alimentation, elles sont pratiquement toutes constituées de rangées de cylindres qui interagissent avec leur environnement fluide. Le mécanisme de capture des particules utilisé dépend des contraintes morphologiques, des besoins énergétiques et des conditions d'écoulement. Comme nos objectifs étaient de comprendre ces relations, nous avons eu recours à des études de comparaison pour interpréter les tendances en nature et pour comprendre les conditions qui provoquent de nouveaux fonctionnements. Nous avons utilisé la dynamique des fluides numérique (computational fluid dynamics, CFD) pour créer des expériences contrôlées et pour simplifier les analyses. Notre première étude démontre que les coûts énergétiques associés au pompage dans les espaces petits sont élevés. De plus, le CFD suggère que les fentes branchiales des ptérobranches sont des structures rudimentaires, d'un ancêtre plus grande. Ce dernier point confirme l'hypothèse qu'un ver se nourrit par filtration tel que l'ancêtre des deuterostomes. Notre deuxième étude détermine la gamme du nombre de Reynolds number critique où la performance d'un filtre de balane change. Quand le Re est très bas, les différences morphologiques n'ont pas un grand effet sur le fonctionnement. Cependant, une pagaie devient une passoire lorsque le Re se trouve entre 1 et 3,5. Le CFD s’est dévoilé être un outil très utile qui a permis d’obtenir des détails sur les microfluides. Ces études montrent comment la morphologie et les dynamiques des fluides interagissent avec la mécanisme de capture ou de structures utilisées, ainsi que comment des petits changements de taille, de forme, ou de vitesse d'écoulement peuvent conduire à un nouveau fonctionnement.

AN ARTIFICIAL NEURAL NETWORK FOR SONAR TARGET DETECTION

Neural Network has emerged as the topic of the day. The spectrum of its application is as wide as from ECG noise filtering to seismic data analysis and from elementary particle detection to electronic music composition. The focal point of the proposed work is an application of a massively parallel connectionist model network for detection of a sonar target. This task is segmented into: (i) generation of training patterns from sea noise that contains radiated noise of a target, for teaching the network;(ii) selection of suitable network topology and learning algorithm and (iii) training of the network and its subsequent testing where the network detects, in unknown patterns applied to it, the presence of the features it has already learned in. A three-layer perceptron using backpropagation learning is initially subjected to a recursive training with example patterns (derived from sea ambient noise with and without the radiated noise of a target). On every presentation, the error in the output of the network is propagated back and the weights and the bias associated with each neuron in the network are modified in proportion to this error measure. During this iterative process, the network converges and extracts the target features which get encoded into its generalized weights and biases.In every unknown pattern that the converged network subsequently confronts with, it searches for the features already learned and outputs an indication for their presence or absence. This capability for target detection is exhibited by the response of the network to various test patterns presented to it.Three network topologies are tried with two variants of backpropagation learning and a grading of the performance of each combination is subsequently made.

Spin filtering through ferromagnetic BiMnO3 tunnel barriers

We report on experiments of spin filtering through ultrathin single-crystal layers of the insulating and ferromagnetic oxide BiMnO3 (BMO). The spin polarization of the electrons tunneling from a gold electrode through BMO is analyzed with a counterelectrode of the half-metallic oxide La2/3Sr1/3MnO3 (LSMO). At 3 K we find a 50% change of the tunnel resistances according to whether the magnetizations of BMO and LSMO are parallel or opposite. This effect corresponds to a spin-filtering efficiency of up to 22%. Our results thus show the potential of complex ferromagnetic insulating oxides for spin filtering and injection.