Marie-Véronique Le Lann

Experimental application of nonlinear model predictive control: temperature control of an industrial semi-batch pilot-plant reactor

This paper describes the application of nonlinear model predictive control (NMPC) to the temperature control of a semi-batch chemical reactor equipped with a multi-fluid heating/cooling system. The strategy of the nonlinear control system is based on a constrained optimisation problem, which is solved repeatedly on-line by a step-wise integration of a nonlinear dynamic model and optimisation strategy. A supervisory control routine has been developed, based on the same nonlinear dynamic model, to handle automatically the fluid changeovers. Both NMPCand supervisory control have been implemented on a PCand applied to a 16 l batch reactor pilot plant. Experiments illustrate the feasibility of such a procedure involving predictive control and supervisory control. # 2002 Elsevier Science Ltd. All rights reserved.

Data validation and missing data reconstruction using self-organizing map for water treatment

Applications in the water treatment domain generally rely on complex sensors located at remote sites. The processing of the corresponding measurements for generating higher-level information such as optimization of coagulation dosing must therefore account for possible sensor failures and imperfect input data. In this paper, self-organizing map (SOM)-based methods are applied to multiparameter data validation and missing data reconstruction in a drinking water treatment. The SOM is a special kind of artificial neural networks that can be used for analysis and visualization of large high-dimensional data sets. It performs both in a nonlinear mapping from a high-dimensional data space to a low-dimensional space aiming to preserve the most important topological and metric relationships of the original data elements and, thus, inherently clusters the data. Combining the SOM results with those obtained by a fuzzy technique that uses marginal adequacy concept to identify the functional states (normal or abnormal), the SOM performances of validation and reconstruction process are tested successfully on the experimental data stemming from a coagulation process involved in drinking water treatment.

Classification Tool Based on Interactivity Between Expertise and Self-Learning Techniques 1

Abstract The present work proposes a learning methodology to identify process situations from measured data with a direct participation of the expert. The classification technique is based in LAMDA fuzzy method of conceptual clustering. The proposed tool has been developed in such a way that the learning phase (Supervised or not) is strongly associated with the expert knowledge. The application chosen for state identification and validation is the Universitat Politecnica de Catalunya fluidised bed gasifier selected as benchmark process in the CHEM European project.

Membership-margin based feature selection for mixed type and high-dimensional data

Definition of membership-margin based objective function.Weighting the antecedents of fuzzy if-then rules to improve classification performance.Evaluate the feature importance within membership-margin framework.Feature selection algorithm for mixed-type and high dimensional data.The effectiveness of the approach is proved through an extensive experimental study. The present paper describes a new feature weighting method based on a membership margin. Distinctive properties of the proposed method include its capability to process problems characterized by mixed-type data (quantitative, qualitative and interval) as well as a huge number of features. The key idea is to map simultaneously all the features of different types into a common space; the membership space. Once all features are represented in a homogeneous space, a feature weighting task can be performed in unified way. This weighting approach is integrated here within a fuzzy classifier through a fuzzy rule weighted concept in order to improve its performance. Each antecedent fuzzy set in the fuzzy if-then rule is weighted to characterize the importance of each proposition and therefore its corresponding feature. Weight estimation process is based on membership margin maximization to estimate a fuzzy weight of each feature in the membership space. Experiments on low and high dimensional real-world datasets demonstrate that the proposed approach can improve significantly the performance of the fuzzy rule-based as well as other state of the art classifiers and can even outperform classical feature weighting approaches. In particular, we show that this approach can yield meaningful results on two real-world applications for cancer prognosis and industrial process diagnosis.

Sensor placement and fault detection using an efficient fuzzy feature selection approach

Process monitoring and fault diagnosis are of great importance for operation safety and efficiency of complex industrial plants. The present article proposes a novel methodology to address the sensor location problem for fault detection. Firstly, all the process situations are identified based on a fuzzy learning algorithm using measurements generated from the whole available set of sensors. Then, a fuzzy feature selection approach is used to select the optimal number of sensors that characterize accurately the set of process situations (abnormal and normal). This method optimizes the performance of the learning algorithm within a membership margin framework, and thereby, it is capable to address correlation and redundancy issues. A behavioral pattern of the process is constructed with the selected sensors and is used to associate new online observations to previously characterized process situations. The proposed strategy has been applied for fault diagnosis to a pharmaceutical synthesis carried out in a new intensified heat-exchanger reactor.

TOWARDS A UNIFIED PRINCIPLE FOR REASONING ABOUT HETEROGENEOUS DATA: A FUZZY LOGIC FRAMEWORK

Human knowledge about monitoring process variables is usually incomplete. To deal with this partial knowledge many types of representation other than the quantitative one are used to describe process variables (qualitative, symbolic interval). Thus, the development of automatic reasoning mechanisms about the process is faced with this problem of multiple data representations. In this paper, a unified principle for reasoning about heterogeneous data is introduced. This principle is based on a simultaneous mapping of data from initially heterogeneous spaces into only one homogeneous space based on a relative measure using appropriate characteristic functions. Once the heterogeneous data are represented in a unified space, a single processing for various analysis purposes can be performed using simple reasoning mechanisms. An application of this principle within a fuzzy logic framework is performed here to demonstrate its effectiveness. We show that simple fuzzy reasoning mechanisms can be used to reason in a unified way about heterogeneous data in three well known machine learning problems.

From Chemical Process Diagnosis to Cancer Prognosis: An Integrated Approach for Diagnosis and Sensor/Marker Selection

Abstract Classification techniques have shown recently their usefulness for complex process diagnosis. Besides the fact that no physical model for the process is required, they enable to study the problem of sensor location. Preliminary studies made previously in the domain of chemical process diagnosis have been the initial key point to extend its application to the medical diagnosis framework. Despite the behavioral difference, both domains exhibit many common practices. However, medical diagnosis recently has brought serious challenges such as high dimensionality (gene expression profiling) and heterogeneity of data (symbolic histo-pathological factors). We show here that both challenges can be overcome and used in return to improve complex process diagnosis.

Symbolic Data Analysis to Defy Low Signal-to-Noise Ratio in Microarray Data for Breast Cancer Prognosis

Microarray profiling has recently generated the hope to gain new insights into breast cancer biology and thereby improve the performance of current prognostic tools. However, it also poses several serious challenges to classical data analysis techniques related to the characteristics of resulting data, mainly high dimensionality and low signal-to-noise ratio. Despite the tremendous research work performed to handle the first challenge in the feature selection framework, very little attention has been directed to address the second one. We propose in this article to address both issues simultaneously based on symbolic data analysis capabilities in order to derive more accurate genetic marker-based prognostic models. In particular, interval data representation is employed to model various uncertainties in microarray measurements. A recent feature selection algorithm that handles symbolic interval data is used then to derive a genetic signature. The predictive value of the derived signature is then assessed by following a rigorous experimental setup and compared with existing prognostic approaches in terms of predictive performance and estimated survival probability. It is shown that the derived signature (GenSym) performs significantly better than other prognostic models, including the 70-gene signature, St. Gallen, and National Institutes of Health criteria.

Use of an Object Oriented Dynamic Hybrid Simulator for the Monitoring of Industrial Processes

PrODHyS is a dynamic hybrid simulation environment, which offers extensible and reusable object oriented components dedicated to the modelling of processes. The purpose of this communication aims at presenting the main concepts of PrODHyS through the modelling and the simulation of a hydraulic system. Then, the feasible use of hybrid simulation in a supervision system is underlined. Copyright

Implementation of model predictive controller in a pharmaceutical development plant

Abstract Predictive control has spread in various domains such as refinery, chemical, metallurgical … industries. Nevertheless, concerning the pharmaceutical industry it still remains relatively exceptional since two particularities are clearly attached to this specific domain: the use of batch processes and the necessity to satisfy to strict validation procedures. In this context, a predictive controller has been developed; the Model Gradient Predictive Controller denoted MGPC and tested in real time application on various chemical reactors in the chemical development plant (PILOT and KILOLAB, Sanofi-Aventis at Montpellier, France). This plant is devoted to investigate new reactions before passing them to an industrial day-to-day production. In such a context, a same apparatus is used for carrying out different operations such as chemical reactions (changing several times a week) and crystallizations with highly non linear temperature set-point profiles (such as cubic profile).