Tid: 8 juni 2006 kl 10.15-11.00 (OBS! Dag och tid)
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Erik Nordblom
Titel: Statistical modelling of sensor response. (Examensarbete)
Sammanfattning: When producing measurement instruments, an important aspect is measurement accuracy which depends upon such factors as environmental conditions, external interfaces and material degradation. Since it may be expensive and time-consuming to conduct an extensive in-field testing procedure, there is good reason to explore effcient test methods.
This thesis presents the development of a general modelling method and how a generated model may be implemented in a calibration procedure for sensors during their operational phase to provide in-field updates to both measurement accuracy and error estimates. As a practical case study of the proposed technique, the response from a pressure sensor is modelled, using only a small amount of data.
The resulting model defines the basic framework of a practical in-field sensor update process. In the specific case of the pressure sensor, the model was shown to be inadequate when used alone but very useful when used as a part of the calibration procedure. The conclusion is that the value of the method is highly dependent upon its implementation and requisite accuracy.