FIGURE 5 - uploaded by Olivier Nerrand
Content may be subject to copyright.
Source publication
INTRODUCTION The development of engineering applications of neural networks makes it necessary to clarify the similarities and differences between the concepts and methods developed for neural networks and those used in more classical fields such as filtering and control. In previous papers [Nerrand et al. 1993], [Marcos et al. 1993], the relations...
Similar publications
The paper proposes a general framework that encompasses the training of neural networks and the adaptation of filters. We show that neural networks can be considered as general nonlinear filters that can be trained adaptively, that is, that can undergo continual training with a possibly infinite number of time-ordered examples. We introduce the can...
Classical adaptive mathematical morphology is based on operators which locally adapt the structuring elements to the image properties. Connected morphological operators act on the level of the flat zones of an image, such that only flat zones are filtered out, and hence the object edges are preserved. Area opening (resp. area closing) is one of the...
Citations
... There are several ways to cope with nonlinear control design for plants subject to uncertainty and disturbances: among them are robust and adaptive control methods. While neural adaptive control is being intensively developed [1] [2] [3] [4] [5], an approach of neural internal model control and of its robustness properties is presented here. In order to precise the scope of this paper, let us first recall basic notions concerning internal model control systems, their design and the properties which make them attractive. ...
... – Simple feedback control systems: their control device is made of the controller only (Fig. 1a). They are the most classic ones in the field of adaptive control (see [1] [2] [3] [5]). However, in the case of non adaptive neural control, since the quality of their performance relies heavily on the fidelity of the model used for the design of the controller, they can only be used if the model is very accurate and if the effect of disturbances is small (see for example [6] [7]). ...
We propose a design procedure of neural internal model control systems for stable processes with delay. We show that the design of such nonadaptive indirect control systems necessitates only the training of the inverse of the model deprived from its delay, and that the presence of the delay thus does not increase the order of the inverse. The controller is then obtained by cascading this inverse with a rallying model which imposes the regulation dynamic behavior and ensures the robustness of the stability. A change in the desired regulation dynamic behavior, or an improvement of the stability, can be obtained by simply tuning the rallying model, without retraining the whole model reference controller. The robustness properties of internal model control systems being obtained when the inverse is perfect, we detail the precautions which must be taken for the training of the inverse so that it is accurate in the whole space visited during operation with the process. In the same spirit, we make an emphasis on neural models affine in the control input, whose perfect inverse is derived without training. The control of simulated processes illustrates the proposed design procedure and the properties of the neural internal model control system for processes without and with delay.
An electroosmotic device was designed to remove water from tomato paste suspensions. The suspension was held in vertical column of 9.1 cm height and between two stainless steel electrodes, an upper serves as anode while the lower serves as a cathode. Direct current electroosmosis dewatering was achieved by the application of an electrical field which allows water to pass through the filter paper held in the lower section. The effects of voltage, current and initial solid concentration on the amount of water removal were examined. The energy of dewatering can be calculated from the measured experimental data. The experimental data were represented by neural network modeling scheme which describes the data accurately and adequately.
Electric vehicle (EV) technologies are a strategic part of research and development in the automotive industry. Among the various kinds of EV prototypes presented by the car manufacturers, fuel cell powered electrical vehicles seem to be a very promising solution. When talking about EV design, a simulation model of the whole fuel cell system is a binding milestone. This would lead in the optimization possibility of the complete vehicle (including all ancillaries, output electrical converter and their dedicated control laws). Nevertheless, the fuel cell system model is strongly dependent of many physico-chemical parameters that are difficult to evaluate on a real proton exchange membrane fuel cell (PEMFC) stack. Moreover, the analytical relations governing the behavior of a PEMFC system are also far from being easy. Thus, a "minimal behavioral model" of a fuel cell system, able to evaluate the output variables and their variations, is highly interesting. Artificial neural networks propose a very efficient tool to reach such an aim. A dynamic recurrent neural network model of a fuel cell system based on proton exchange membrane technology is presented in this paper.
Among the various kinds of electrical vehicle (EV) prototypes presented by the car manufacturers, fuel-cell EVs seem to be a very promising solution. Five different fuel-cell technologies are available in the research laboratories. Nevertheless, only two technologies can really be considered for transportation applications due to their solid electrolyte, i.e., proton exchange membrane fuel cells (PEMFCs) and solid oxide fuel cells. The PEMFCs are investigated in this paper. When talking about EV design, a simulation model of the whole fuel-cell system is a binding milestone. This would lead in the optimization ability of the complete vehicle (including all ancillaries, output electrical converter, and their dedicated control laws). Nevertheless, the fuel-cell model is strongly dependent on many physicochemical parameters that are difficult to evaluate on a real PEMFC stack. Moreover, the analytical relations governing the behavior of a PEMFC system are also far from being easy. Thus, a ldquominimal behavioral modelrdquo of a fuel-cell system, which is able to evaluate the output variables and their variations, is highly interesting. Artificial neural networks propose a very efficient tool to reach such an aim. In this paper, a PEMFC neural network model is proposed.
