CONTRIBUTION A LA COMMANDE PREDICTIVE OPTIMALE SOUS CONTRAINTES DES PROCEDES DISCONTINUS NON LINEAIRES UTILISES DANS L'INDUSTRIE ALIMENTAIRE

Abstract : The main goal of this thesis is the elaboration and the experimental validation of an automatic control method, well suited for nonlinear, slow and open loop stable batch processes, frequently encountered in food industry. Batch processes operate in unsteady state, and have no setpoints defined for the relevant variables. Our method allows real time computation of the control variables' trajectories which optimize the global operation cost, taking into account the process dynamics and technological and quality constraints. A state and parameter estimator uses the information provided by on line measurements, introducing output feedback. The future control strategy is adjusted in real time, based on the predictions of the internal model. The predictions are made in a long time perspective, typically up to the end of the batch. Our method does not require any particular form of the cost function, constraints, internal model or state estimator. All components may be nonlinear and designed separately. The experimental validation is made on two pilot plants. The application to the drying process shows that with a suitable choice of the internal model and of the control optimization algorithm, the complexity of the on line computations can be kept very low. The application to the refrigeration process uses an internal model based on differential and algebraic equations, together with a relatively sophisticated optimization algorithm, requiring a standard desktop computer. Both applications show the ability of our algorithm to fulfill the control specifications (cost reduction and constraint satisfaction), even in the presence of strong disturbances, both external and parametric, measured and unmeasured.
Document type :
Theses
Complete list of metadatas

Cited literature [128 references]  Display  Hide  Download

https://hal-agroparistech.archives-ouvertes.fr/tel-01537229
Contributor : Ioan-Cristian Trelea <>
Submitted on : Monday, June 12, 2017 - 3:07:27 PM
Last modification on : Tuesday, June 13, 2017 - 1:10:52 AM
Long-term archiving on : Thursday, September 14, 2017 - 12:40:43 PM

Identifiers

  • HAL Id : tel-01537229, version 1

Collections

Citation

Ioan-Cristian Trelea. CONTRIBUTION A LA COMMANDE PREDICTIVE OPTIMALE SOUS CONTRAINTES DES PROCEDES DISCONTINUS NON LINEAIRES UTILISES DANS L'INDUSTRIE ALIMENTAIRE. Sciences de l'ingénieur [physics]. Ecole Nationale Supérieure des Industries Agricoles et Alimentaires - ENSIA, 1997. Français. ⟨tel-01537229⟩

Share

Metrics

Record views

169

Files downloads

829