Existing techniques for sensor validation and sensor fusion are often based on analytical sensor models. Such models can be arbitrarily complex and consequently Gaussian distributions are often assumed, generally with a detrimental effect on overall system performance. A holistic approach has therefore been adopted in order to develop two novel and complementary approaches to sensor validation and fusion based on empirical data. The first uses the Nadaraya-Watson kernel estimator to provide competitive sensor fusion. The new algorithm is shown to reliably detect and compensate for bias errors, spike errors, hardover faults, drift faults and erratic operation, affecting up to three of the five sensors in the array. The inherent smoothing action of the kernel estimator provides effective noise cancellation and the fused result is more accurate than the single 'best sensor'. A Genetic Algorithm has been used to optimise the Nadaraya-Watson fuser design. The second approach uses analytical redundancy to provide the on-line sensor status output μH∈[0,1], where μH=1 indicates the sensor output is valid and μH=0 when the sensor has failed. This fuzzy measure is derived from change detection parameters based on spectral analysis of the sensor output signal. The validation scheme can reliably detect a wide range of sensor fault conditions. An appropriate context dependent fusion operator can then be used to perform competitive, cooperative or complementary sensor fusion, with a status output from the fuser providing a useful qualitative indication of the status of the sensors used to derive the fused result. The operation of both schemes is illustrated using data obtained from an array of thick film metal oxide pH sensor electrodes. An ideal pH electrode will sense only the activity of hydrogen ions, however the selectivity of the metal oxide device is worse than the conventional glass electrode. The use of sensor fusion can therefore reduce measurement uncertainty by combining readings from multiple pH sensors having complementary responses. The array can be conveniently fabricated by screen printing sensors using different metal oxides onto a single substrate.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

This thesis discusses a general approach to hybrid powertrain control based on optimisation and optimal control techniques. A typical strategy comprises a high level non-linear control for optimised energy efficiency, and a lower level Linear Quadratic Regulator (LQR) to track the high-level demand signals and minimise the first torsional vibration mode. The approach is demonstrated in simulation using a model of the Toyota Prius hybrid vehicle, and comparisons are made with a simpler control system which uses proportional integral (PI) control at the lower level. The powertrain of the Toyota Prius has a parallel configuration, comprising a motor, engine and generator connected via an epicyclic gear train. High level control is determined by a Power Efficient Controller (PE C) which dynamically varies the operating demands for the motor, engine and generator. The PEC is an integrated nonlinear controller based on an iterative downhill search strategy for optimising energy efficiency and battery state of charge criteria, and fully accounts for the non-linear nature of the various efficiency maps. The PEC demand signals are passed onto the LQR controller where a cost function balances the importance of deviations from these demands against an additional criterion relating to the amplitude of driveline vibrations. System non-linearity is again accounted for at the lower level through gain scheduling of the LQR controller. Controller performance is assessed. in simulation, the results being compared with a reference system that uses simple PI action to deliver low-level control. Consideration is also given to assessing performance against that of a more general, fully non-linear dynamic optimal controller.

This thesis considers the application of neural networks to automotive suspension systems. In particular their ability to learn non-linear feedback control relationships. The speed of processing, once trained, means that neural networks open up new opportunities and allow increased complexity in the control strategies employed. The suitability of neural networks for this task is demonstrated here using multilayer perceptron, (MLP) feed forward neural networks applied to a quarter vehicle simulation model. Initially neural networks are trained from a training data set created using a non-linear optimal control strategy, the complexity of which prohibits its direct use. They are shown to be successful in learning the relationship between the current system states and the optimal control.