Recurrent Trainable Neural Network Research Articles

A direct adaptive neural control scheme with integral terms

A direct adaptive neural control scheme with single and double integral-plus-state (IPS) actions is proposed. The control scheme contains two recurrent trainable neural network (RTNN) models, which are a plant parameter identifier and state estimator, an IPS feedback/feedforward controller, and one or two I-terms. The good performance of the adaptive IPS control scheme is confirmed by closed-loop systems analysis and by simulation results obtained with a MIMO plant, corrupted by noise. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 213–224, 2005.

Adaptive recurrent neural network control of biological wastewater treatment

Three adaptive neural network control structures to regulate a biological wastewater treatment process are introduced: indirect, inverse model, and direct adaptive neural control. The objective is to keep the concentration of the recycled biomass proportional to the influent flow rate in the presence of periodically acting disturbances, process parameter variations, and measurement noise. This is achieved by the so-called Jordan Canonical Recurrent Trainable Neural Network, which is a completely parallel and parametric neural structure, permitting the use of the obtained parameters, during the learning phase, directly for control system design. Comparative simulation results confirmed the applicability of the proposed control schemes. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 173–193, 2005.

AN ADAPTIVE NEURAL CONTROL SYSTEM OF A DC MOTOR DRIVE.

An improved parallel recurrent neural network with canonical architecture, named Recurrent Trainable Neural Network (RTNN), and an error based back-propagation through time learning algorithm, are applied to a D.C. motor drive identification and control. The unknown nonlinear dynamics of the motor together with the load are identified by the RTNN. The trained RTNN identifier is combined with a desired reference model and a RTNN controller in a direct adaptive control scheme, so in order to achive a desired trajectory tracking of the motor speed and position. Finally, the applicability of the RTNN topology and learning is illustrated by experimental results.

A Recurrent Neural Network Model of the B.t. Fed-Batch Fermentation Process

This paper proposes the use of a new Recurrent Trainable Neural Network model (RTNN) for modeling the fed-batch fermentation kinetics data of Bacillus thuringiensis as a function of the operational conditions. The proposed RTNN model has ten inputs, six outputs and sixteen neurons in the hidden layer, with global and local feedbacks. The learning algorithm is a modified version of the backpropagation through time. The approximation error is below 2% and the generalization error is below 6%. The learning was performed in 101 epochs, 92 iterations each one. The RTNN model validation was done with experimental data not included in the learning process.