This paper presents an investigation into the development of parametric and non-parametric approaches for dynamic modelling of a flexible manipulator system. The least mean squares, recursive least squares and genetic algorithms are used to obtain linear parametric models of the system. Moreover, non-parametric models of the system are developed using a non-linear AutoRegressive process with eXogeneous input model structure with multi-layered perceptron and radial basis function neural networks. The system is in each case modelled from the input torque to hub-angle, hub-velocity and end-point acceleration outputs. The models are validated using several validation tests. Finally, a comparative assessment of the approaches used is presented and discussed in terms of accuracy, efficiency and estimation of the vibration modes of the system.

This study develops a neural network system to monitor the safety of a bridge structure. A truck of constant mass is driven at constant speed through the target bridge. Then, the maximal and minimal values of the bridge elongations are processed by a monitoring system to evaluate the current condition of the bridge. The monitoring system is composed of parallel backpropagation neural networks. Each neural network monitors a part of the bridge. The neural networks are trained using simulation data. The numerical example shows that the monitoring system is effective in the damage detection of the bridge.

The effects of different dendritic geometries and distal dendritic Na+ current distributions on the propagation of action potentials (APs) and sensory EPSPs were investigated using a multi-compartment model of thalamocortical (TC) neurones where the somatic and proximal dendritic distribution of voltage-gated channels matched the ones measured experimentally, i.e. a uniform distribution of K+ currents and a non-uniform distribution of Na+ and T-type Ca2+ currents.

Our simulations indicated that the distal dendritic Na+ channel density has not to be larger than 50% of the somatic density in order to reproduce the electrical activities recorded experimentally from the soma and proximal dendrites of TC neurones. Moreover, we could highlight the existence of a distinct threshold density of distal dendritic Na+ channels necessary to support the regeneration of APs in this part of the dendritic tree: this threshold density was smaller for non-branching than for heavily branching dendrites.

The amplitude of the somatic EPSP mainly depended on the number of simultaneously activated synapses on any dendritic branch, despite large differences in the size of the dendritic EPSPs. The amplitude of the EPSP on a proximal dendrite was also dependent on the number and relative location of simultaneously activated synapses on all other proximal dendritic branches. The dendritic geometry did not affect these features of the simulated sensory EPSPs. In addition, the duration of somatic and proximal dendritic EPSPs was markedly increased (100%) in the presence of somatic and proximal dendritic T-type Ca2+ current.

The backpropagation of EPSPs to distal dendrites was affected by the dendritic Na+ channel distributions, but even in the absence of distal dendritic Na+ channels the EPSP reached the dendritic ends with less than 40% decrease in amplitude. Overall, the amplitude of the backpropagating EPSP was not greatly affected by the dendritic geometry, though a smaller amplitude reduction in unbranched than in heavily branching dendrites was observed.

This paper describes the design of an Internal Model Control (IMC) system for a planar two-degree-of-freedom robot. IMC was investigated as an alternative to the basic inverse control scheme which is difficult to implement. The proposed IMC system consisted of a forward internal neural model of the robot, a neural controller and a conventional feedback controller, all of which were realised easily. Both the neural model and the neural controller were based on recurrent networks which were trained using the backpropagation (BP) algorithm. The paper presents the results obtained with two types of recurrent networks as well as a conventional PID system.