MATHMOD 2015 featured four plenary lectures:
- What has Instrumental Variable method to offer for system identification?
Prof. Marion Gilson-Bagrel
Université de Lorraine, Nancy, France - Model-based approaches for the future sustainable aircraft: the EU ADDSAFE project
Dr. Philippe Goupil, R&T Engineer
AIRBUS – Flight Control System, Toulouse, France - Advanced Mechatronics for Precision Engineering and Mechatronic Imaging Systems
Prof. Georg Schitter
Vienna University of Technology, Austria - (Dynamic) Iteration Schemes for Coupled Problems in Electrical Engineering
Prof. Sebastian Schöps
TU Darmstadt, Germany
Abstracts
What has Instrumental Variable method to offer for system identification?
Prof. Marion Gilson-Bagrel
Université de Lorraine, Nancy, France
For many years, several attempts have been made to identify accurate linear or non linear plant models, by several algorithms, and with or without the need to identify full noise models.
More precisely, when looking at methods that can consistently identify plant models of systems while relying on simple linear (regression) algorithms, instrumental variable (IV) techniques seem to be rather attractive, but at the same time not very often applied.
However, IV methods of parameter estimation have a long history in the statistical and control engineering literature and present some interesting properties.
On the other hand, when dealing with highly complex processes that are high dimensional in terms of inputs and outputs, it can be attractive to rely on methods that do not require non-convex optimization algorithms. Besides this computationally attractive property, IV methods have the potential advantage that they can identify plant models consistently when the noise model is misspecified, and in the particular case of closed-loop system identification, when the present controller is non-linear and/or time-varying.
The aim of this talk is to give
some clues of the use of those IV methods in several kinds of
applications (linear, LPV, time domain, frequency domain, environmental
data set, water quality measurements…) and to illustrate what IV methods
can offer to system identification.
Model-based approaches for the future sustainable aircraft: the EU ADDSAFE project
Dr. Philippe Goupil, R&T Engineer
AIRBUS – Flight Control System, Toulouse, France
In this presentation a high-fidelity aircraft benchmark, developed by Airbus for advanced flight control related studies, is exposed in relation to its evolution towards advanced fault diagnosis within a European 7th Framework Program project entitled “Advanced Fault Diagnosis for Sustainable Flight Guidance and Control (ADDSAFE)”. This European project was established to study and facilitate the transfer of model-based fault detection and diagnosis methods from Academia to Industry. The importance of the project arose on the one hand, due to the representativeness of the benchmark, and on the other hand, the industrial benchmarking and validation of the developed designs. The project had the overall aim of researching and developing model-based methods for detecting and isolating aircraft flight control system faults: predominantly sensor and actuator malfunctions. Most of the model-based methods rely on the idea of analytical redundancy in which, in contrast to physical or hardware redundancy, real physical measurements are complemented with analytically computed redundant variables. A common method to analytically detect the existence of a failure is to look for anomalies in the plant’s output relative to a model-based estimate of that output, generating a so-called residual. The generated residual has to include enough information to determine that a specific fault has occurred.
This presentation provides details on the ADDSAFE project motivation and objectives. It describes the aircraft model and the fault scenarios defined by Airbus and which constitutes the industrial benchmark. A cursory overview of the model-based methods studied during the project is proposed. Finally, an industrial perspective is given especially by focusing on the golden rules to be followed to help bridge the gap between basic research levels and industrial needs.
Advanced Mechatronics for Precision Engineering and Mechatronic Imaging Systems
Prof. Georg Schitter
Vienna University of Technology, Austria
Mechatronic imaging systems, such as atomic force microscopes (AFM), wafer scanners, adaptive optics, and laser scanning microscopes, demand a continuous improvement of system speed, range, and precision, which requires advanced mechatronic designs and highly sophisticated motion control.
Modeling and simulation are key enabling technologies for the development and engineering of mechatronic systems in the high-tech industry. Already at the system design phase all components involved in the specific application have to be considered, where a well predictable behaviour of the system components is required. Examples for these components are the mechanical structure of the device, the power amplifier, the actuators, the sensors, electronics, and the real-time control system. To meet the demanding specifications, the final system, including all hard- and software components, has to be tailored to and optimized for each specific application.
This presentation addresses these challenges by illustrating examples for precision motion control, AFM imaging, confocal laser scanning microscopy and adaptive optics. The presented examples successfully demonstrate the potential to significantly improve the performance of mechatronic imaging systems via an integrated mechatronic design approach by utilizing the interplay between process design and control design.
(Dynamic) Iteration Schemes for Coupled Problems in Electrical Engineering
Prof. Sebastian Schöps
TU Darmstadt, Germany
Today, due to increased accuracy of modeling and simulation, multi-domain problems become more and more important in many engineering applications. Often a monolithic approach, i.e., the solution of all subproblems in one system, is cumbersome or even impossible because incompatible algorithms or software packages are involved. Thus simulation engineers need to “weakly” couple subproblems in an efficient and stable way, such that each problem can be tackled separately. For example for time-dependent problems where different time scales are present, waveform relaxation schemes are a promising approach that allows for an efficient simulation of the problem. However, the independent treatment introduces a splitting error, which should be mitigated by iterative procedures that in turn can cause computational overhead. In this talk we discuss theoretical and practical issues as existence and uniqueness, accuracy, stability and numerical efficiency of the schemes and addressed advantages and disadvantages for electromagnetic applications.