Data interpretation model-based methods

Among nonlocal methods, those based on linear projection are the most widely used for data interpretation. Owing to their limited modeling ability, linear univariate and multivariate methods are used mainly to extract the most relevant features and reduce data dimensionality. Nonlinear methods often are used to directly map the numerical inputs to the symbolic outputs, but require careful attention to avoid arbitrary extrapolation because of their global nature. 

With the view that a KBS interpreter is a method for mapping from input data in the form of intermediate symbolic state descriptions to labels of interest, four families of approaches are described here, each offering inference mechanisms and related knowledge representations that can be used to solve interpretation problems namely, model-based approaches, digraphs, fault trees, and tables. These methods have been heavily used... 

A fixed target value for a is preferable and can be arrived at in several ways. It could be fixed arbitrarily, with a value based on a perception of how laboratories should perform. It could be an estimate of the precision required for a specific task of data interpretation, a could be derived from a model of precision, such as the Horwitz Curve .15 However, while this model provides a general picture of reproducibility, substantial deviation from it may be experienced for particular methods. 

Some researchers have doubted whether nonporphyrin metal compounds exist at all in petroleum (Goulon et al., 1984 Berthe et al., 1984). This speculation is based on the similarities in spectral features of petroleum-bound metals and model-metal porphyrin compounds from selected petroleums under the examination of x-ray absorption, photoelectron, and laser desorption spectroscopies. Interpretation of spectra from these methods, however, is uncertain and difficult given the lack of sufficient data on model compounds, particularly nonporphyrin metals and in a petroleum environment or matrix. 

Localized quasi-linear inversion increases the accuracy and efficiency of wave-field data interpretation because it is based on a much more accurate forward modeling solution than the Born approximation, used in the original Bleistein method. An example of successful application of the localized QL approximation in radar-diffraction tomography can be found in (Zhou and Liu, 2000). 

The severe computational burden associated with assembling and carrying out adsorption calculations on disordered model microstructures for porous solids, such as those discussed in Sections ILA and II.B, has until recently limited the development of pore volume characterization methods in this direction. While the reahsm of these models is highly appealing, their application to experimental isotherm or scattering data for interpretation of adsorbent pore structure remains cumbersome due to the structural complexity of the models and the computational resources that must be brought to bear in their utilization. Consequently, approximate pore structure models, based upon simple pore shapes such as shts or cylinders, have been retained in popular use for pore volume characterization. 

Thermogravimetry is an attractive experimental technique for investigations of the thermal reactions of a wide range of initially solid or liquid substances, under controlled conditions of temperature and atmosphere. TG measurements probably provide more accurate kinetic (m, t, T) values than most other alternative laboratory methods available for the wide range of rate processes that involve a mass loss. The popularity of the method is due to the versatility and reliability of the apparatus, which provides results rapidly and is capable of automation. However, there have been relatively few critical studies of the accuracy, reproducibility, reliability, etc. of TG data based on quantitative comparisons with measurements made for the same reaction by alternative techniques, such as DTA, DSC, and EGA. One such comparison is by Brown et al. (69,70). This study of kinetic results obtained by different experimental methods contrasts with the often-reported use of multiple mathematical methods to calculate, from the same data, the kinetic model, rate equation g(a) = kt (29), the Arrhenius parameters, etc. In practice, the use of complementary kinetic observations, based on different measurable parameters of the chemical change occurring, provides a more secure foundation for kinetic data interpretation and formulation of a mechanism than multiple kinetic analyses based on a single set of experimental data. 

---