A mixture of experts network structure construction algorithm for modelling and control

This paper introduces a new fast, effective and practical model structure construction algorithm for a mixture of experts network system utilising only process data. The algorithm is based on a novel forward constrained regression procedure. Given a full set of the experts as potential model bases, the structure construction algorithm, formed on the forward constrained regression procedure, selects the most significant model base one by one so as to minimise the overall system approximation error at each iteration, while the gate parameters in the mixture of experts network system are accordingly adjusted so as to satisfy the convex constraints required in the derivation of the forward constrained regression procedure. The procedure continues until a proper system model is constructed that utilises some or all of the experts. A pruning algorithm of the consequent mixture of experts network system is also derived to generate an overall parsimonious construction algorithm. Numerical examples are provided to demonstrate the effectiveness of the new algorithms. The mixture of experts network framework can be applied to a wide variety of applications ranging from multiple model controller synthesis to multi-sensor data fusion.

Neurofuzzy design and model construction of nonlinear dynamical processes from data

A common problem in many data based modelling algorithms such as associative memory networks is the problem of the curse of dimensionality. In this paper, a new two-stage neurofuzzy system design and construction algorithm (NeuDeC) for nonlinear dynamical processes is introduced to effectively tackle this problem. A new simple preprocessing method is initially derived and applied to reduce the rule base, followed by a fine model detection process based on the reduced rule set by using forward orthogonal least squares model structure detection. In both stages, new A-optimality experimental design-based criteria we used. In the preprocessing stage, a lower bound of the A-optimality design criterion is derived and applied as a subset selection metric, but in the later stage, the A-optimality design criterion is incorporated into a new composite cost function that minimises model prediction error as well as penalises the model parameter variance. The utilisation of NeuDeC leads to unbiased model parameters with low parameter variance and the additional benefit of a parsimonious model structure. Numerical examples are included to demonstrate the effectiveness of this new modelling approach for high dimensional inputs.

Synchronisation error resistant CDMA detection

A bit-level linear CDMA detector is presented which is based on the minimum variance distortionless response (MVDR) principle. Owing to the interference suppression capability made possible by basing the detector on the MVDR principle and the fact that no inversion of the user correlation matrix is involved, the influence of synchronisation errors is greatly suppressed.

Synchronisation error resistant multiuser detection for asynchronous CDMA

Linear CDMA detectors have emerged as a promising solution to multiple access interference (MAI) suppression. Unfortunately, most existing linear detectors suffer from high sensitivity to synchronisation errors (also termed parameter estimation error), and synchronisation error resistant detectors have so far not been as widely investigated as they should have. This paper extends the minimum variance distortionless response (MVDR) detector, proposed previously by this author (Zheng 2000) for synchronous systems, to asynchronous systems. It has been shown that the MVDR structure is equally effective for asynchronous systems, especially for the weaker users.

A moving-mesh finite element method and its application to the numerical solution of phase-change problems

A distributed Lagrangian moving-mesh finite element method is applied to problems involving changes of phase. The algorithm uses a distributed conservation principle to determine nodal mesh velocities, which are then used to move the nodes. The nodal values are obtained from an ALE (Arbitrary Lagrangian-Eulerian) equation, which represents a generalization of the original algorithm presented in Applied Numerical Mathematics, 54:450--469 (2005). Having described the details of the generalized algorithm it is validated on two test cases from the original paper and is then applied to one-phase and, for the first time, two-phase Stefan problems in one and two space dimensions, paying particular attention to the implementation of the interface boundary conditions. Results are presented to demonstrate the accuracy and the effectiveness of the method, including comparisons against analytical solutions where available.

Numerical modeling of inhaled charged aerosol deposition in human airways

A new numerical modeling of inhaled charge aerosol has been developed based on a modified Weibel's model. Both the velocity profiles (slug and parabolic flows) and the particle distributions (uniform and parabolic distributions) have been considered. Inhaled particles are modeled as a dilute dispersed phase flow in which the particle motion is controlled by fluid force and external forces acting on particles. This numerical study extends the previous numerical studies by considering both space- and image-charge forces. Because of the complex computation of interacting forces due to space-charge effect, the particle-mesh (PM) method is selected to calculate these forces. In the PM technique, the charges of all particles are assigned to the space-charge field mesh, for calculating charge density. The Poisson's equation of the electrostatic potential is then solved, and the electrostatic force acting on individual particle is interpolated. It is assumed that there is no effect of humidity on charged particles. The results show that many significant factors also affect the deposition, such as the volume of particle cloud, the velocity profile and the particle distribution. This study allows a better understanding of electrostatic mechanism of aerosol transport and deposition in human airways.

Equalisation with adaptive time lag

In an adaptive equaliser, the time lag is an important parameter that significantly influences the performance. Only with the optimum time lag that corresponds to the best minimum-mean-square-error (MMSE) performance, can there be best use of the available resources. Many designs, however, choose the time lag either based on preassumption of the channel or simply based on average experience. The relation between the MMSE performance and the time lag is investigated using a new interpretation of the MMSE equaliser, and then a novel adaptive time lag algorithm is proposed based on gradient search. The proposed algorithm can converge to the optimum time lag in the mean and is verified by the numerical simulations provided.

Forecast impact of targeted observations: sensitivity to observation error and proximity to steep orography

For a targeted observations case, the dependence of the size of the forecast impact on the targeted dropsonde observation error in the data assimilation is assessed. The targeted observations were made in the lee of Greenland; the dependence of the impact on the proximity of the observations to the Greenland coast is also investigated. Experiments were conducted using the Met Office Unified Model (MetUM), over a limited-area domain at 24-km grid spacing, with a four-dimensional variational data assimilation (4D-Var) scheme. Reducing the operational dropsonde observation errors by one-half increases the maximum forecast improvement from 5% to 7%–10%, measured in terms of total energy. However, the largest impact is seen by replacing two dropsondes on the Greenland coast with two farther from the steep orography; this increases the maximum forecast improvement from 5% to 18% for an 18-h forecast (using operational observation errors). Forecast degradation caused by two dropsonde observations on the Greenland coast is shown to arise from spreading of data by the background errors up the steep slope of Greenland. Removing boundary layer data from these dropsondes reduces the forecast degradation, but it is only a partial solution to this problem. Although only from one case study, these results suggest that observations positioned within a correlation length scale of steep orography may degrade the forecast through the anomalous upslope spreading of analysis increments along terrain-following model levels.

The distribution of solar wind speeds during solar minimum: calibration for numerical solar wind modeling constraints on the source of the slow solar wind

It took the solar polar passage of Ulysses in the early 1990s to establish the global structure of the solar wind speed during solar minimum. However, it remains unclear if the solar wind is composed of two distinct populations of solar wind from different sources (e.g., closed loops which open up to produce the slow solar wind) or if the fast and slow solar wind rely on the superradial expansion of the magnetic field to account for the observed solar wind speed variation. We investigate the solar wind in the inner corona using the Wang-Sheeley-Arge (WSA) coronal model incorporating a new empirical magnetic topology–velocity relationship calibrated for use at 0.1 AU. In this study the empirical solar wind speed relationship was determined by using Helios perihelion observations, along with results from Riley et al. (2003) and Schwadron et al. (2005) as constraints. The new relationship was tested by using it to drive the ENLIL 3-D MHD solar wind model and obtain solar wind parameters at Earth (1.0 AU) and Ulysses (1.4 AU). The improvements in speed, its variability, and the occurrence of high-speed enhancements provide confidence that the new velocity relationship better determines the solar wind speed in the outer corona (0.1 AU). An analysis of this improved velocity field within the WSA model suggests the existence of two distinct mechanisms of the solar wind generation, one for fast and one for slow solar wind, implying that a combination of present theories may be necessary to explain solar wind observations.

Ensemble prediction for nowcasting with a convection-permitting model - II: forecast error statistics

A 24-member ensemble of 1-h high-resolution forecasts over the Southern United Kingdom is used to study short-range forecast error statistics. The initial conditions are found from perturbations from an ensemble transform Kalman filter. Forecasts from this system are assumed to lie within the bounds of forecast error of an operational forecast system. Although noisy, this system is capable of producing physically reasonable statistics which are analysed and compared to statistics implied from a variational assimilation system. The variances for temperature errors for instance show structures that reflect convective activity. Some variables, notably potential temperature and specific humidity perturbations, have autocorrelation functions that deviate from 3-D isotropy at the convective-scale (horizontal scales less than 10 km). Other variables, notably the velocity potential for horizontal divergence perturbations, maintain 3-D isotropy at all scales. Geostrophic and hydrostatic balances are studied by examining correlations between terms in the divergence and vertical momentum equations respectively. Both balances are found to decay as the horizontal scale decreases. It is estimated that geostrophic balance becomes less important at scales smaller than 75 km, and hydrostatic balance becomes less important at scales smaller than 35 km, although more work is required to validate these findings. The implications of these results for high-resolution data assimilation are discussed.