Fabrice Druaux

• Assistant Professor
• Université Le Havre Normandie

This paper presents a multi-objective indirect neural adaptive control design for nonlinear square multi-variable systems with unknown dynamics. The control scheme is made of an adaptive instantaneous neural emulator, a neural controller based on fully connected real-time recurrent learning (RTRL) networks and an online parameter updating law. A mu...

In this study, an adaptive control based on fuzzy adapting rate for neural emulator of nonlinear systems having unknown dynamics is proposed. The indirect adaptive control scheme is composed by the neural emulator and the neural controller which are connected by an autonomous algorithm inspired from the real-time recurrent learning. In order to ens...

The present study concerns the remote monitoring of immersed plate-like structures as the ones used for marine current turbines. The innovation of this work is the remote damage detection based on a systematic analysis of a small set of ultrasonic measurements limited by the backscattered echoes from the structure edges. The detection and localizat...

In this paper, multimodel and neural emulators are proposed for uncoupled multivariable nonlinear plants with unknown dynamics. The contributions of this paper are to extend the emulators to multivariable non square systems and to propose a systematic method to compute the multimodel synthesis parameters. The effectiveness of the proposed emulators...

Non-destructive testing and structural health monitoring are essential for safety and reliability of marine kinetic energy related fields. In this work we address the problem of damage detection of underwater finite length plates using acoustic inverse scattering and image processing methods. Time series signals and the 2D images obtained from thes...

This paper deals with the adaptive control of single-input multi-output (SIMO) underactuated nonlinear systems. The restriction of the control authority for these systems causes major difficulties in control design. In this work, we propose an adaptive neural controller based on neural emulator to solve the control problems for a class of SIMO nonl...

Detection and diagnosis method is proposed for surface damage in immersed structures. It is based on noncontact ultrasonic echography measurements, signal processing tools, and artificial intelligence methods. Significant features are extracted from the measured signals and a classification method is developed to detect the echoes resulting from su...

This paper investigates robust adaptive control for unknown non-linear and multivariable systems with fully connected recurrent neural networks. On-line weights updating law and closed loop performance are derived from the Lyapunov approach. Robust stability under the parametric uncertainties due to disturbances of the overall system is provided. T...

The developed method is based on acoustic underwater scattering measurements and can be applied to characterize a defect in a blade of a marine current turbine. To simplify the study, the blade is replaced by a rectangular plate immersed in water having a groove opening out. The measurements are made on horizontal plane perpendicular to the long ax...

Health monitoring is investigated for immersed structures. The environment of these structures makes their monitoring and diagnosis very difficult. In this paper, the major challenge is to make easy and efficient the monitoring of this kind of structures. The proposed detection method is based on non contact measurements with acoustic scattering. I...

This work describes multivariable adaptive neural control based on multimodel emulator for nonlinear square MIMO systems. Multimodel approach is an interesting alternative and a powerful tool for modelling and emulating complex processes. This paper deals with the identification of nonlinear MIMO systems employing an uncoupled multimodel. Efficienc...

The proposed method concerns the detection and diagnosis of surface faults in immersed structures. This method is suitable for the monitoring of inaccessible systems as stream turbine systems whose unavailability causes serious economic consequences. It is based on an active diagnosis by surface acoustic wave propagation. The generation and acquisi...

In this paper, multimodel and neural emulators are proposed for nonlinear plants with unknown dynamics. The contribution of this paper is to extend the emulators to multi-variable non square systems. The effectiveness of the proposed emulators are shown through an illustrative simulation example. The obtained results are very satisfactory and show...

The immersed system environment makes delicate their diagnosis. Such systems encountered in energy production sites (stream turbines), water treatment or oil platform are not easily accessible for diagnosis issues. Most of dedicated diagnosis techniques are expensive, need a precise positioning of sensors and present large delay. The need to develo...

Le contrôle par contact est le plus répandu dans l'inspection des systèmes mais son application exige un accès facile au système. La nécessité de développer une technique de contrôle sans contact pour des systèmes difficiles d'accès est donc justifiée. La contribution proposée s'inscrit dans cette démarche pour le diagnostic de défaut des systèmes...

This paper investigates adaptive control design for nonlinear square MIMO systems. The control scheme is based on recurrent neural networks emulator and controller with decoupled adaptive rates. Networks' parameters are updated according to an autonomous algorithm inspired from the Real Time Recurrent Learning (RTRL). The contributions of this pape...

This paper adresses a Lyapunov stability analysis of nonlinear systems control. We consider an adaptive control scheme based on recurrent neural networks emulator and controller with decoupled adaptive rates. Lyapunov sufficient stability conditions for decoupled adaptive rates of the emulator and controller are proposed. In order to guarantee the...

Neural networks have been widely used in the field of intelligent information processing such as classification, clustering, prediction, and recognition. In this paper, a non-parametric supervised classifier based on neural networks is proposed for diagnosis issues. A parameter selection with self adaptive growing neural network (SAGNN) is develope...

Detecting and diagnosing faults before deteriorating the system performance is a crucial task for the reliability and safety of many engineering systems. A parameter selection with Self Adaptive Growing Neural Network (SAGNN) is developed for automatic Fault Detection and Diagnosis (FDD) in industrial environments. The growing and adaptive skill of...

This work investigates an uncoupled multimodel emulator for non-linear system control design. Efficiency of the proposed multimodel emulator is illustrated by comparison with the neural one by their application to SISO indirect adaptive neural control. Neither an initialisation parameter and nor online adaptation is required for multimodel emulator...

This paper is about the control design of hybrid dynamical systems modelled with Petri nets. For this purpose, continuous Petri nets with variable speeds are investigated and described as piecewise bilinear linear state space representations. In this context, the marking vector is considered as a state space vector, subsets of places are defined as...

Fault detection and diagnosis have gained widespread industrial interest in machine monitoring due to their potential advantage that results from reducing maintenance costs, improving productivity and increasing machine availability. This article develops an adaptive intelligent technique based on artificial neural networks combined with advanced s...

For the improvement of reliability, safety and efficiency advanced methods of supervision, fault detection and fault diagnosis become increasingly important for many technical processes. To determine the condition of an inaccessible fault in an operating mechanical system, the vibration signal of the machine is continuously monitored by placing sen...

In this paper, we develop an indirect adaptive control struc- ture based on recurrent neural networks. An adaptive emulator inspired from the Real Time recurrent Learning algorithm is presented. Neural network does not learn the plant dynamics but emulates the input-output mapping with a small time window. Thereafter, a controller with a struc- tur...

Early detection and isolation of faults can help avoid major system breakdowns. This paper presents a non parametric fault diagnosis method to detect and isolate faults in industrial environment. The proposed Input Output Classification Mapping (IOCM) algorithm is based on mapping the input parameters from Gaussian hidden layer functions of an RBF...

This paper provides an adaptation algorithm for the control of complex system via recurrent neural networks. The proposed method is derived from RTRL algorithm. Neural emulator and neural controller parameters are one-line updated independently. To illustrate the tracking and the disturbance rejection capabilities of the real time control algorithm...

To maintain an efficient operating unit and avoid failure of mineral critical equipment, fault detection and diagnosis are the solution for the critical parts of these equipments. This paper presents a non parametric classification technique using Best Selective Parameters (BSP) embedded in Binary Decision Tree (BDT). The method arises from the que...

This paper deals with a new indirect adaptive control scheme with decoupled adaptive rates, developed for complex square systems with unknown dynamics. This scheme, based on fully connected neural networks, is inspired from the real time recurrent learning (RTRL) algorithm. Both neural emulator and neural controller networks do not learn the plant...

In this paper, a real time recurrent learning-based emulator is presented for nonlinear plants with unknown dynamics. This emulator is based on fully connected recurrent neural networks. Starting from zero values, updating rate, time parameter and weights of the instantaneous neural emulator adapt themselves in order to estimate the process output....

The objective of our work is to detect and isolate the faults that occur in large scale systems with many measurements. An advanced fault detection and isolation (FDI) method is proposed based on wavelet decomposition, Parameters Elimination Method (PEM) and Radial Basis Function (RBF) networks. Input signals are decomposed into approximations (low...

An autonomous indirect scheme is proposed for multivariable process control and is extended to unstable open-loop plant-wide processes. Our principal objective in this work is to prove the feasibility to control an industrial plant by a small size neural system without any a priori training. The control scheme is made of an adaptive instantaneous n...

In this paper, stable indirect adaptive control with recurrent neural networks is presented for square multivariable non-linear plants with unknown dynamics. The control scheme is made of an adaptive instantaneous neural model, a neural controller based on fully connected “Real-Time Recurrent Learning” (RTRL) networks and an online parameters updat...

A feed-forward neural network is proposed for monitoring operating modes of large scale processes. A Gaussian hidden layer associated with a Kohonen output layer map the principal features of measurements of state variables. Subsets of selective neurons are generated into the hidden layer by means of self adapting of centers and dispersions paramet...

Adaptive control by means of neural networks for non-linear dynamical systems is an open issue. For real world applications, practitioners have to pay attention to external disturbances, parameters uncertainty and measurement noise, as long as these factors will influence the stability of the closed loop system. As a consequence, robust stability o...

Petri nets (PN) are useful for the modelling, analysis and control of hybrid dynamical systems (HDS) because PN combine in a comprehensive way discrete events and continuous behaviours. On one hand, PN are suitable for modelling the discrete part of HDS and for providing a discrete abstraction of continuous behaviours. On the other hand, continuous...

This paper investigates robust adaptive control for unknown nonlinear systems with fully connected recurrent neural networks. On-line weights updating law and closed loop performance are derived from the Lyapunov approach. Mathematical proof for the robust stability under the parametric uncertainties due to disturbances of the overall system is pro...

In this paper fully connected RTRL neural networks are studied. In order to learn dynamical behaviours of linear-processes or to predict time series, an autonomous learning algorithm has been developed. The originality of this method consists of the gradient based adaptation of the learning rate and time parameter of the neurons using a small pertu...

The selection of learning rates to obtain satisfactory performances for neural network controllers is a challenging problem. In order to skip any time consuming experimentation for the choice of an appropriate value of the learning rate, this paper is concerned with an online adaptive learning rate algorithm derived from the convergence analysis of...

This paper is about the flow control design of hybrid dynamical systems modelled with continuous Petri nets and described as a set of bilinear state space representations. The marking vector stands for the state space vector, where some places are observable and other are not. The transitions are divided into controllable ones and uncontrollable on...

in this paper, stable indirect adaptive control with recurrent neural networks is presented for multi-input multi-output (MIMO) square non linear plants with unknown dynamics. The control scheme is made of a neural model and a neural controller based on fully connected RTRL networks. On-line weights updating law, closed loop performance, and bounde...

Cet article concerne la commande des systèmes dynamiques hybrides modélisés par réseaux de Petri continus. Les réseaux de Petri utilisés sont décrits sous forme de représentations d'état : le vecteur d'état correspond au vecteur de marquage, le vecteur de sortie est défini par des sommes pondérées de sous ensembles de places, et le vecteur d'entrée...

This paper is about traffic flow short term prediction and monitoring based on magnetic sensors measurements. For these purposes, the advantages and drawbacks of feed-forward and real time recurrent learning neural networks are investigated. Structures determination, weights initialization, networks training and automatic incidents detection are di...

The shape of detrital quartz grains, mainly acquired under the effect of mechanical transport agents, has been observed and described for a long time by sedimentologists. The roughness of these grains is recognized as an important parameter for analyzing sediments, but to quantify this parameter with a discriminant measure can be difficult. By usin...

This paper is about the control design of hybrid dynamical systems modeled with Petri nets. For this purpose, continuous Petri nets with variable speeds are described as state space representations. In this context, the marking vector is considered as a state space vector, the model outputs are defined by subnets, and the transitions are divided in...

This paper is about the control design of hybrid dynamical systems modeled with Petri nets. For this purpose, continuous Petri nets with variable speeds are described as state space representations. In this context, the marking vector is considered as a state space vector, the model outputs are defined by subnets, and the transitions are divided in...

In this paper, a multiscale roughness descriptor for particles is proposed. This descriptor, based on the harmonic wavelet transform, acts as a mathematical microscope and allows analyses of wear and erosion phenomena acting on particles. The roughness descriptor is applied to a problem of sediments analysis and classification and its performances...

By using the concept of multiresolution analysis and the wavelet transform, we develop a new shape descriptor for sedimentary particles which allows the analysis of their roughness at different scales. The performance of the method is tested on a problem of sediment classification and compared to the fractal dimension.

A fully connected continuous time recurrent neural network, trained by means of Real-Time Recurrent Learning, is investigated. A theoretical analysis of the output vector of the network during the training stage is performed. We point out the necessity to apply an additional constraint to the synaptic weight matrix with the intention of reducing th...

Some theoretical results related to the behavior of a large collection of neurons composed of two and three families are presented. Each of them is represented by the mean value of the characteristic parameters such as the potential membrane and weights of the connections. Description in the state space has been retained. Attention is given to the...

The dynamic behavior (comportment) of an artificial neural network depends not only on the synaptic weights, but also on the stimuli input. Without stimuli, the network stable state is permanent or oscillatory. The external stimuli input modifies the neural state, changing it from one stable permanent state to another or from an oscillatory to a pe...

Using r.f. sputtering we have fabricated non-crystalline thin films of NiAg alloys, two metals which are insoluble in both liquid and solid states. The variation in resistivity with composition exhibits a maximum. The Faber-Ziman theory often used to explain both resistivity and temperature coefficient behaviours needs the measurement of interferen...

Ni0.5Ag0.5 thin films were prepared by R.F. sputtering. The magnetic properties (saturation magnetization and susceptibility) was measured by means of a Faraday balance in the temperature range 170–715K. The films are made up of fine particles of pure nickel scattered in an amorphous matrix. Below 400K the films are paramagnetic and above they have...