A Recursive Frequency Estimator Using Linear Prediction and a Kalman-Filter-Based Iterative Algorithm

This paper proposes a new Kalman-filter-based recursive frequency estimator for discrete-time multicomponent sinusoidal signals whose frequencies may be time-varying. The frequency estimator is based on the linear prediction approach and it employs the Kalman filter to track the linear prediction coefficients (LPCs) recursively. Frequencies of the sinusoids can then be computed using the estimated LPCs. Due to the coloredness of the linear prediction error, an iterative algorithm is employed to estimate the covariance matrix of the prediction error and the LPCs alternately in the Kalman filter in order to improve the tracking performance. Simulation results show that the proposed Kalman-filter-based iterative frequency estimator can achieve better tracking results than the conventional recursive least-squares-based estimators.     

Onset of buoyancy-driven convection in melting from below

When a horizontal homogeneous solid is melted from below, convection can be induced in a thermally unstable melt layer. In this study the onset of buoyancy-driven convection during time-dependent melting is investigated by using similarly transformed disturbance equations. The critical Rayleigh numbers based on the melt-layer thickness are found numerically for various conditions. For small superheats, the present predictions approach the well known results of classical Rayleigh-Bénard problems, that is, critical Rayleigh numbers are located between 1,296 and 1,708, regardless of the Prandtl number. However, for high superheats the critical Rayleigh number increases with an increase in phase change rate but with decrease in Prandtl number.     

Cyclic oxidation response of multiphase niobium-based alloys

Cyclic oxidation tests were performed on multiphase Nb-based alloys containing silicide, Laves, and Nb solid solution phases. In particular, the oxidation resistance of six alloys with various compositions (Nb, Ti, Hf, Cr, Ge, and Si) and microstructures was characterized by thermal cycling from ambient temperature to a peak temperature that ranges from 900 °C to 1400 °C. Weight change data were obtained and the corresponding spalled oxides were collected and identified by X-ray diffraction. The results indicated that Nb-based alloys formed a mixture of CrNbO₄, Nb₂O₅, and Nb₂O₅ · TiO₂, with possibly small amounts of SiO₂ or GeO₂. The oxidation resistance was improved when CrNbO₄ formed instead of Nb₂O₅ and Nb₂O₅ · TiO₂. These results were used to assess the influence of microstructure and composition on the oxidation resistance of multiphase Nb-based alloys.     

Ice-slurry production using direct contact heat transfer

An ice slurry generation system was developed using direct contact heat transfer between water and the coolant, Fluroinert FC-84. The location of the coolant nozzle is an important design consideration to avoid clogging due to freezing of water. An ice fraction of up to about 40 percent was obtained with the nozzle located at the bottom of the ice slurry tank and the jet directed upwards into the water. Two simplified model were developed to extract the heat transfer coefficient between the coolant drops and the water. The first model requires as input the average drop diameter and the residence time while the second model uses the measured drop diameter distribution. The estimated heat transfer coefficients are much smaller than those computed using single-sphere correlations.     

Infections in patients with cystic fibrosis following lung transplantation

BACKGROUND: There is controversy over whether colonization with drug-resistant organisms is a contraindication to lung transplantation. METHODS: We undertook a retrospective review of the results of lung transplantation for patients with cystic fibrosis (CF) at Duke University Medical Center. RESULTS: As of May 1996, 21 patients with CF underwent bilateral lung transplantation. The first patient died within 24 h of transplantation from sepsis due to Stenotrophomonas maltophilia. Of the remaining 20 patients, 17 (85%) are alive and in stable condition. The three deaths were related primarily to bronchiolitis obliterans at 4 and 18 months in two patients and to cytomegalovirus pneumonitis at 5 months in the other patient. The 17 surviving patients have been followed up for a mean of 13 months (range, 0.5 to 34 months). Most of them were colonized and infected with multidrug-resistant organisms before transplantation. Following transplantation, 11 patients had complications from infections. One patient had bacteremia due to a panresistant Burkholderia cepacia and was treated successfully. Two patients had bacteremia and wound infection due to Burkholderia gladioli, previously thought to be pathogenic only in plants. Both patients were treated successfully. Of the six patients with Aspergillus fumigatus isolated from cultures before transplantation, only one had invasive disease following transplantation and responded to treatment. CONCLUSION: The organisms present before transplantation were not the primary cause of mortality in our patient population. Our findings suggest that lung transplantation should be considered in CF patients infected with multidrug-resistant organisms.     

A new method to estimate parameters of linear compartmental models using artificial neural networks

At present, the preferred tool for parameter estimation in compartmental analysis is an iterative procedure; weighted nonlinear regression. For a large number of applications, observed data can be fitted to sums of exponentials whose parameters are directly related to the rate constants/coefficients of the compartmental models. Since weighted nonlinear regression often has to be repeated for many different data sets, the process of fitting data from compartmental systems can be very time consuming. Furthermore the minimization routine often converges to a local (as opposed to global) minimum. In this paper, we examine the possibility of using artificial neural networks instead of weighted nonlinear regression in order to estimate model parameters. We train simple feed-forward neural networks to produce as outputs the parameter values of a given model when kinetic data are fed to the networks' input layer. The artificial neural networks produce unbiased estimates and are orders of magnitude faster than regression algorithms. At noise levels typical of many real applications, the neural networks are found to produce lower variance estimates than weighted nonlinear regression in the estimation of parameters from mono- and biexponential models. These results are primarily due to the inability of weighted nonlinear regression to converge. These results establish that artificial neural networks are powerful tools for estimating parameters for simple compartmental models.