Research Topics

My research activity concerns process monitoring and supervision and more precisely the development or improvement of model-based fault detection methods and the design of fault tolerant control laws. That kind of problem can be treated with two different approaches. The first is based on the knowledge of an analytical model of the behavior of the system to monitor (model-based methods); the second uses only system input and output records and the analysis is carried out using data analysis methods (data-driven methods). Regarding the first approach, the models used may be more or less complex (linear models or models linearized around particular operating points, nonlinear models, especially Takagi-Sugeno models or linear parameter varying models) taking into account possible uncertainties and unknown inputs. The second approach relies on conventional and / or classification data analysis methods: principal component analysis, support vector machines and mainly their extension to the nonlinear case using kernels.

Keywords:
Fault detection, fault isolation, fault estimation, model-based methods, observer design, state and/or parameter estimation, fault tolerant control, Takagi-Sugeno models, LPV or quasi-LPV models, reconfiguration, data-driven methods, PCA, SVM, classification method, kernel methods

---

Ongoing PhD

Krishnan Srinivasarengan -- Contribution to the optimization and monitoring of energy management of non-residential buildings

Start date: october 2014.
European project "Energy in Time".
Main Supervisors: Christophe Aubrun & Didier Maquin

This study is part of European Project FP7 "Simulation-based control for Energy Efficiency building operation and maintenance: Energy in time" which begun in October 2013 and last 4 years.

Buildings operational stage represents 80% of building's life-cycle cost of which 50% is consequence of the energy use. Moreover, up to 90% of the buildings' life cycle carbon emissions occur during their operational phase, mainly as consequence of the HVAC (Heating, Ventilation and Air Conditioning), lighting and appliances' energy use1. Therefore, energy and cost saving strategies addressing this building operation phase will have a major impact in the building life cycle cost.

The buildings' energy demand and consumption is influenced by numerous factors both inherent and external to buildings. Aspects such as the constructive characteristics, climate, building usage or users' behaviour, among others, directly affect the final energy performance. Furthermore, it is important to highlight that a good energy design of the building does not implies to obtain a good energy performance of the building. The integration of adequate control and operational strategies is a must in order buildings reach to their optimal efficiency levels.

Usually, the objective is to optimally schedule and balance the power flow between energy sources and storage, based on operational and lifecycle metrics (e.g. energy, cost, and emissions), customer preferences, and operational constraints and continuously adjust HVAC (Heating, Ventilation and Air-Conditioning) set-points in order to adapt the building energy performance to occupancy and weather loads, based on the actual and forecast information. This will allow minimizing HVAC energy consumption and peak demand while maintaining the indoor environment within users expected comfort conditions. This can be achieved under the assumption that all equipments operate properly. When equipment degradation or operation faults affect the building performance and service quality, some Fault Detection and Diagnostics (FDD) modules together with Fault-Tolerant Controls (FTC) need to be integrated within the building systems, directly interacting with the Building Energy Management System (BEMS). Thus, through the effective combination and interaction of these tools, the faults effect will be maintained within an acceptable range from the desired system behaviour variations.

The objective of the thesis is to develop new Fault-Tolerant Control approaches for BEMS based on Model Predictive Control (MPC) principle. MPC relies on building model as well as on weather forecasts and occupancy predictions in order to find the optimal control sequence to be implemented in the future. Only the first element in the sequence is actually applied to the building. At the next decision instant, a new optimal control sequence is computed, and so on. The best control sequence is found by solving, at each moment in the decision process, an on line optimization problem. MPC's ability to handle constrained multivariable systems, performances constraints as well as economic objectives makes this paradigm particularly well suited for the issue of energy of FTC buildings. The integrated fault identification and fault-tolerant control strategy to be designed must be able to handle nonlinear behaviour of energetic systems. Within this context, different kinds of buildings' model will be investigated among which Takagi-Sugeno technique aiming at representing the nonlinear behaviour by the means of time-varying weighted sum of linear models.

Mahjoub El Mountassir -- Design of data-driven structural health monitoring methods. Application to tubular structures

Start date: may 2015.
CIFRE with l'Institut de Soudure (site de Yutz)
Main supervisor: Didier Maquin (with Gilles Mourot)

Kwami Dodzivi Anani -- Fault isolation and estimation using kernel principal component analysis

Start date: october 2015.
Public contract with IAEM Doctoral School
Main supervisor: Didier Maquin (with Maya Kallas)

Top