Use of kernel-based Bayesian models to predict late osteolysis after hip replacement

We studied the relationship between osteolysis and polyethylene wear, age at surgery, body mass index and height in 463 subjects (180 osteolysis and 283 controls) after cemented Charnley total hip arthroplasty (THA), in order to develop a kernel-based Bayesian model to quantitate risk of osteolysis. Such tools may be integrated into decision-making algorithms to help personalize clinical decision-making. A predictive model was constructed, and the estimated posterior probability of the implant failure calculated. Annual wear provided the greatest discriminatory information. Age at surgery provided additional predictive information and was added to the model. Body mass index and height did not contain valuable discriminatory information over the range in which observations were densely sampled. The robustness and misclassification rate of the predictive model was evaluated by a five-times cross-validation method. This yielded a 70% correct predictive classification of subjects into osteolysis versus non-osteolysis groups at a mean of 11 years after THA. Finally, the data were divided into male and female subsets to further explore the relationship between wear rate, age at surgery and incidence of osteolysis. The correct classification rate using age and wear rate in the model was approximately 66% for males and 74% for females.     

Nonlinear system identification using wavelet networks

A novel identification scheme using wavelet networks is presented for nonlinear dynamical systems. Based on fixed wavelet networks, parameter adaptation laws are developed using a Lyapunov synthesis approach. This guarantees the stability of the overall identification scheme and the convergence of both the parameters and the state errors, even in the presence of modelling errors. Using the decomposition and reconstruction techniques of multiresolution decompositions, variable wavelet networks are introduced to achieve a desired estimation accuracy and a suitable sized network, and to adapt to variations of the characteristics and operating points in nonlinear systems. B-spline wavelets are used to form the wavelet networks and the identification scheme is illustrated using a simulated example.     

Computational system identification for Bayesian NARMAX modelling

In this contribution we derive a computational Bayesian approach to NARMAX model identification. The identification algorithm exploits continuing advances in computational processing power to numerically obtain posterior distributions for both model structure and parameters via sampling methods. The main advantage of this approach over other NARMAX identification algorithms is that for the first time model uncertainty is characterised as a byproduct of the identification procedure. The algorithm is based on the reversible jump Markov chain Monte Carlo (RJMCMC) procedure. Key features of the approach are (i) sampling of unselected model terms for testing for inclusion in the model (the birth move), which encourages global searching of the model term space, (ii) sampling of previously selected model terms for testing for exclusion from the model—a naturally incorporated pruning step (the death move), which leads to model parsimony, and (iii) estimation of model and parameter distributions, which are naturally generated in the Bayesian framework. We present a numerical example to demonstrate the algorithm and a comparison with a forward regression method: the results show that the RJMCMC approach is competitive and gives useful additional information regarding uncertainty in both model parameters and structure.     

Dynamic model tracking design for low inertia, high speed permanent magnet ac motors

Permanent magnet ac (PMAC) motors have existed in various configurations for many years. The advent of rare-earth magnets and their associated highly elevated levels of magnetic flux makes the permanent magnet motor attractive for many high performance applications from computer disk drives to all electric racing cars. The use of batteries as a prime storage element carries a cost penalty in terms of the unladen weight of the vehicle. Minimizing this cost function requires the minimum electric motor size and weight to be specified, while still retaining acceptable levels of output torque. This tradeoff can be achieved by applying a technique known as flux weakening which will be investigated in this paper. The technique allows the speed range of a PMAC motor to be greatly increased, giving a constant power range of more than 4:1. A dynamic model reference controller is presented which has advantages in ease of implementation, and is particularly suited to dynamic low inertia applications such as clutchless gear changing in high performance electric vehicles. The benefits of this approach are to maximize the torque speed envelope of the motor, particularly advantageous when considering low inertia operation. The controller is examined experimentally, confirming the predicted performance.     

System identification from multiple short-time-duration signals

System identification problems often arise where the only modeling records available consist of multiple short-time-duration signals. This motivates the development of a modeling approach that is tailored for this situation. An identification algorithm is presented here for parameter estimation based on minimizing the simulated prediction error, across multiple signals. The additional complexity of estimating the initial states corresponding to each signal is removed from the estimation algorithm. A numerical simulation demonstrates that the proposed algorithm performs well in comparison to the often-used least squares method (which leads to biased estimates when identifying systems from measurement noise corrupted signals). The approach is applied to the identification of the passive oculomotor plant; parameters are estimated that describe the dynamics of the plant, which represent the time constants of the visco-elastic elements that characterize the plant connective tissue.     

An iterative Kalman smoother/least-squares algorithm for the identification of delta-ARX models

Additive measurement noise on the output signal is a significant problem in the δ-domain and disrupts parameter estimation of auto-regressive exogenous (ARX) models. This article deals with the identification of δ-domain linear time-invariant models of ARX structure (i.e. driven by known input signals and additive process noise) by using an iterative identification scheme, where the output is also corrupted by additive measurement noise. The identification proceeds by mapping the ARX model into a canonical state-space framework, where the states are the measurement noise-free values of the underlying variables. A consequence of this mapping is that the original parameter estimation task becomes one of both a state and parameter estimation problem. The algorithm steps between state estimation using a Kalman smoother and parameter estimation using least squares. This approach is advantageous as it avoids directly differencing the noise-corrupted ‘raw’ signals for use in the estimation phase and uses different techniques to the common parametric low-pass filters in the literature. Results of the algorithm applied to a simulation test problem as well as a real-world problem are given, and show that the algorithm converges quite rapidly and with accurate results.     

Fault detection and isolation in non-linear stochastic systems—A combined adaptive Monte Carlo filtering and likelihood ratio approach

This paper presents the development of a new method for solving fault detection and isolation (FDI) problem in general non-linear stochastic systems. In this paper, the faults are modelled as unknown changes in system parameters and adaptive Monte Carlo filtering approach is used in deriving an FDI scheme. Essentially, a set of adaptive Monte Carlo filters are designed based on the augmented system models along with a nominal Monte Carlo filter designed based on the nominal system model. The likelihood functions of the observations are then evaluated using the particles from these (adaptive) Monte Carlo filters and FDI is eventually achieved via the likelihood ratio test. The simulation results on a highly non-linear system are provided which demonstrates the effectiveness of the proposed method.