Sensor Signal Analysis

Data analysis is used to detect interesting information in measured data, build models that explains dependencies and behaviour of all kinds of systems, and is an integrated part of most measurement systems. SINTEF uses a wide range of methods to analyse and model sensor signals, from mechanistic models obtained from a physical/chemical understanding of the system, via semi-empirical statistical methods, to advanced non-linear multivariate methods with fuzzy decisions.

Mechanistic models are relevant and always preferred when a good physical/chemical understanding of the system is available. Only a limited number of sensor measurements may then be sufficient to derive a desired system parameter, detect or filter noise in the data, or detect deviations from desired system states.

One example where physical models play an important role is in optics where for instance optical diffraction modelling and full electromagnetic field modelling can be used with high accuracy in the design and evaluation of optical systems and in the interpretation of measured signals.

On the other extreme, purely empirical methods build mathematical models of systems based on observed data only. To obtain reliable results such methods generally require large amounts of data that span the full range of operation. These methods are thus most relevant when the mechanistic understanding is limited, but where the behaviour of the systems can easily be measured during normal or stimulated variations.

In many cases we do not have clear cut situations as above, but need to combine a limited mechanistic understanding of the system with a limited possibility for measurements. We should then use methods that utilize both sources of information optimally. A large range of methods can be used, from a simple linearization of a known instrument non-linearity, to the use of local regime and ASMOD models, artificial neural networks and fuzzy logic.

Application areas:

• Gas measurement. Gas concentration or identification is the basis for a number of process control or safety applications. SINTEF ICT has wide experience in gas sensing. This activity is two-fold: on one hand we focus on high performance gas sensing, on the other the focus is on practical (often low cost) single gas sensors.
• Recycling technology. The green-wave slogan ”Waste is displaced resources” is certainly true. However, recycling works best where resource recovery can be made into profitable industry. The key to profitable recycling lies in reliable and efficient classification and separation technology.
• Instrument calibration
• Smart sensors
• Industrial process instrumentation and monitoring
• Modelling of biomedical, social and economic systems
• Image analysis

Infrared light is the key to a sorting machine for drink cartons and plastic bottles developed by SINTEF.