Interpretation by measuring, modelling and identification III

The IMMI project is focused on the development of measurement techniques (measuring, modelling, identification) and measurement equipment to be used to gather the required information.
Instrumentation evolved towards a common architecture: an analogue front end, analogue to digital converter, signal processing hardware and some capability to connect a computer network. Because of this evolution modular measurement systems became reality (VME, VXI, MMS, PC cards...). The technical elements dominating in general the success of instruments are:
good analogue front ends and signal conditioners, closely related to the application area;
fast digital components;
the capability and quality of the algorithms used for the information processing, the correction of systematic errors, and the possibility to link the raw measurements with a model of the device under test.
For a long time the principal research of the department was directed towards the third element in this list. The first IMMI project (1989-1994) of the department described the measurement problem as an identification problem (experiment design, estimation algorithms, expert system to guide the inexperienced user). In the second project the development and application of identification methods to measurement problems continued but also the building of dedicated instruments started. Because the department has neither the infrastructure, nor the experience to built good analogue front ends and the experience to develop fast digital components, cooperation with instrument manufactures to develop these instruments was initiated. Amongst them it is worth mentioning AGILENT TECHNOLOGIES (NMDG-Brussels, advanced measurement techniques: building of a calibrated microwave network analyser for nonlinear systems), SEBA Service (building of an advanced reflectometer to characterize 'copper' based transmission lines), and IMEC (Interuniversitair microelectronica centrum). At this moment, prototypes of the network analyser and the reflectometer are operational. Also the identification techniques developed in the department are distributed on a large scale (frequency domain identification toolbox distributed as such by the Mathworks).

Aim and planning of the IMMI project

The IMMI project can be divided in three layers. The first layer consists of basic measurement techniques which strongly depend on the application field. The equipment used to measure the properties of a microwave device is completely different from the instruments used to analyse a mechanical system. It also includes the organization of the instrument setup, and its integration in a computer network. It is obvious that there is a strong interaction between the basic instrument structure and the organization of a complete measurement setup. In this project work a pulsed microwave network analyser for nonlinear systems will be developed and the efforts to build a reflectometer for the characterization of communication lines will continue.
The second layer consists of the identification/signal processing techniques used to extract the information of interest from the raw data. These methods should minimize the influence of measurement errors (for example noise) on the final results. They are the link between measuring and information technology and are already less dependent on the specific application. In the second layer projects such information extraction methods are developed and applied to a wide variety of problems (modelling linear lumped and distributed systems, measuring non linear systems in the mechanical and the micro-wave field), always using identification theories as a principal guideline. Ideally, the result should not only be an identified model but should also provide an error bound on it, quantifying the impact of stochastic errors (for example noise) and modelling errors (for example the impact of nonlinear distortions on a linear model).
The third layer adds the intelligence to the measurement system. Modern measurement techniques (both the instruments and the identification algorithms) are very powerful if they are used correctly. However the required competence for a correct use is rapidly increasing. Also the consequences of incorrectly use of the modern equipment become more severe because an inexperienced user even doesn't notice that his results are useless due to wrong manipulations of his sophisticated apparatus. To avoid that the equipment can only be used by specialized technicians and to maintain access to the most recent technical applications it will be necessary to implement expert knowledge in the instrument itself to guide the user to a correct solution of his problem.
Intelligence, on the other hand, can be required also when indirect measurements of physical quantities are at stake, which do not fall into the category of sensors/transducers, because the behaviour of the latter can be modelled, calibrated and so forth using the schemes of the second IMMI layer. To illustrate the to be added intelligence in indirect measurement procedures the characterization of a sea floor can be mentioned. The properties of bottom sediments in a geotechnical sense can be estimated indirectly using an underwater acoustical remote sensing technique. Using the first two IMMI layers, estimates of the acoustical parameters of sediments can be produced. Using the third IMMI layer will allow the generation of geotechnical estimates of sediments in a wide sense when intelligence is added to the system. This approach involves expert system techniques, based on database methods, because ambiguities arise due to the non unique one to one mapping of acoustic and sedimentological parameter sets.
From experience it turned out that over the years, algorithms and techniques are pushed down from the upper layers to the lower layers. The instruments themselves become smarter (for example a DFT (FFT) analysis becomes part of the front end of the instrument), and some expert knowledge is pushed down to the identification techniques by using more robust algorithms (for example automatic model selection in the identification process). This process will continue in the future by developing more robust algorithms, which perform complicated tasks automatically (for example automatic noise analysis). In these steps a good theoretic understanding is indispensable in order to understand the impact of simplification and even to generate new algorithms. For this reason a lot of attention will be paid on the robustification and the user aspects of the algorithms so that expert knowledge can be pushed down into the second layer. The research projects of IMMI are mainly organized along the first and the second layer. However, for the projects which aim at monitoring systems where a lot of total different measured and estimated parameters are required, e.g. environmental monitoring by modelling, the interpretation of the observations require interdisciplinary expertise, and hence, expert systems still will be proposed.

first layer projects
measurement and modelling of microwave systems
development of a pulsed microwave analyser for (non)linear systems;
development of a reflectometer for the characterization of communication lines.
second layer projects
The projects are grouped into three classes:
development and robustification of identification algorithms and making them more user friendly;
characterization of media: materials using underwater acoustics and transmission lines using reflectometry;
development of measurement/calibration techniques for micro-wave analysers and the modelling of micro-wave components from measurements and simulations.

third layer projects
building of functional relationships between estimated acoustic parameters obtained by acoustic remote sensing of a sea floor and the geotechnical and sedimentological properties of the sediments (both of those in suspension and of the first meters of the sea bottom);
prediction of the behaviour of complex non stationary systems from the interpretation of large data sets;
development of auto-consistent software for complex measurement systems.