Skip to main content
eScholarship
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

The Instrumental Model

Abstract

The goal of this dissertation is to enable better predictive models by engaging raw experimental data through the Instrumental Model. The Instrumental Model captures the protocols and procedures of experimental data analysis. The approach is formalized by encoding the Instrumental Model in an XML record. Decoupling the raw experimental data from the data analysis procedure, the Instrumental Model provides means for rigorous uncertainty quantification of predictive model.

The concept of the Instrumental Model and its data model, which governs how the data is described, is discussed in this work. The Instrumental Model XML record is linked to raw experimental data records and calibration data records, providing a complete description of the experimental data.

The Instrumental Model approach is first demonstrated using a set of formaldehyde oxidation shock-tube experiments. In those experiments, the transmitted laser light intensity was measured by a photodiode and the produced voltages were recorded by a computer. The corresponding Instrumental Model transforms these raw data into CO concentration upon the user's request.

The Instrumental Model is expanded by performing uncertainty quantification of model predictions using raw data from a more complex experiment - a stoichiometric C2H2/O2/Ar premixed laminar flame mapped with VUV-photoionization molecular-beam mass spectrometry. The experimental signals were modeled with a premixed laminar flame code augmented with an Instrumental Model, designed to link raw signals to derived properties. The consistency of the model and raw experimental data are quantified and predictions for weak-signal observations of O, OH, C2H3 and unknown background H2O mole fractions are made.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View