A priori

Finally, values of n must be obtained when specific ("chemical") interactions can occur. These are difficult to estimate a priori but reasonable approximations can often be made by choosing a value (Appendix C) for a chemically similar system. 

ZT,ZJ Temporary true vapor composition estimates. k6 True vapor composition estimate flag (0 = a priori ... 

For homogeneous NDT data and repeatable inspection conditions successful automated interpretation systems can relatively easily be developed. They usually use standard techniques from statistical classification or artificial intelligence. Design of successful automated interpretation systems for heterogeneous data coming form non-repeatable, small volume inspections with little a-priori information about the pieces or constructions to be inspected is far more difficult. This paper presents an approach which can be used to develop such systems. 

Neural network classifiers. The neural network or other statistical classifiers impose strong requirements on the data and the inspection, however, when these are fulfilled then good fully automatic classification systems can be developed within a short period of time. This is for example the case if the inspection is a part of a manufacturing process, where the inspected pieces and the possible defect mechanisms are well known and the whole NDT inspection is done in repeatable conditions. In such cases it is possible to collect (or manufacture) as set of defect pieces, which can be used to obtain a training set. There are some commercially available tools (like ICEPAK [Chan, et al., 1988]) which can construct classifiers without any a-priori information, based only on the training sets of data. One has, however, always to remember about the limitations of this technique, otherwise serious misclassifications may go unnoticed. 

When implementing CBR systems one has to able to define and implement the methods to distinguish between data from different classes. This is a more difficult problem than when constructing a simple data classifier, as the important parameters cannot be simply determined based on a set of examples. One has to have some a-priori knowledge about the important features that distinguish various data classes, as well as anticipate possible data forms that can be encountered during future inspections. This may necessitate the use of more features to describe the problem than a comparable classifier would use. When determining the data... 

The classical computer tomography (CT), including the medical one, has already been demonstrated its efficiency in many practical applications. At the same time, the request of the all-round survey of the object, which is usually unattainable, makes it important to find alternative approaches with less rigid restrictions to the number of projections and accessible views for observation. In the last time, it was understood that one effective way to withstand the extreme lack of data is to introduce a priori knowledge based upon classical inverse theory (including Maximum Entropy Method (MEM)) of the solution of ill-posed problems [1-6]. As shown in [6] for objects with binary structure, the necessary number of projections to get the quality of image restoration compared to that of CT using multistep reconstruction (MSR) method did not exceed seven and eould be reduced even further. 

The starting point of this approach is that the 3D restoration is implemented by the solution of the variational problem for the trade-off functional M , which favors in a weighted manner measured data (functional A) and a priori knowledge (functional B) ... 

In the Maximum Entropy Method (MEM) which proceeds the maximization of the conditional probability P(fl p ) (6) yielding the most probable solution, the probability P(p) introducing the a priory knowledge is issued from so called ergodic situations in many applications for image restoration [1]. That means, that the a priori probabilities of all microscopic configurations p are all the same. It yields to the well known form of the functional 5(/2 ) [9] ... 

We will see that superseding the functional fi(p ) in the form of Gibbs measure (4) ensures the linearity of equation (1), simplifies the iteration procedure, and naturally provides the support of any expected feature in the image. The price for this is, that the a priori information is introduced in more biased, but quite natural form. 

Often the a priori knowledge about the structure of the object under restoration consists of the knowledge that it contains two or more different materials or phases of one material. Then, the problem of phase division having measured data is quite actual. To explain the mathematical formulation of this information let us consider the matrix material with binary structure and consider the following potentials ... 

The importance of distinct a priori knowledge account becomes more perceptible if noisy data are under restoration. The noise / ( shifts the solution of (1) from the Maximum Likelihood (ML) to the so called Default Model for which the function of the image constraint becomes more significant. 

R. A. Roberts Limited data tomography using support minimization with a priori data. Review of Progress in QNDE, vol. 11, ed. by D. Thompson and D. Chimenti, Plenum Press, New York, 1992, pp. 549-555. 

V. Vengrinovich, Yu. Denkevich, G.-R. Tillack, and S. Heine X-ray 3-D Reconstruction Using minimal Projections and Maximum a Priori Knowledge. Proc. of Int. Conf CM NDT 95, Novemb. 21-24, Minsk, pp.77-81, (1995). 

No a priori information about the unknown profile is used in this algorithm, and the initial profile to start the iterative process is chosen as (z) = 1. Moreover, the solution of the forward problem at each iteration can be obtained with the use of the scattering matrices concept [8] instead of a numerical solution of the Riccati equation (4). This allows to perform reconstruction in a few seconds of a microcomputer time. The whole algorithm can be summarized as follows ... 

We assmne that the invariance group is known a priori. 

There is some uncertainty connected with testing techniques, errors of characteristic measurements, and influence of fectors that carmot be taken into account for building up a model. As these factors cannot be evaluated a priori and their combination can bring unpredictable influence on the testing results it is possible to represent them as additional noise action [4], Such an approach allows to describe the material and testing as a united model — dynamic mathematical model. 

There have been numerous efforts to inspect specimens by ultrasonic reflectivity (or pulse-echo) measurements. In these inspections ultrasonic reflectivity is often used to observe changes in the acoustical impedance, and from this observation to localize defects in the specimen. However, the term defect is related to any discontinuity within the specimen and, consequently, more information is needed than only ultrasonic reflectivity to define the discontinuity as a defect. This information may be provided by three-dimensional ultrasonic reflection tomography and a priori knowledge about the specimen (e.g., the specimen fabrication process, its design, the intended purpose and the material). A more comprehensive review of defect characterization and related nondestructive evaluation (NDE) methods is provided elsewhere [1]. 

The a priori information involved by this modified Beta law (5) does not consider the local correlation between pixels, however, the image f is mainly constituted from locally constant patches. Therefore, this a priori knowledge can be introduced by means of a piecewise continuous function, the weak membrane [2]. The energy related to this a priori model is ... 

This criterion resumes all the a priori knowledge that we are able to convey concerning the physical aspect of the flawed region. Unfortunately, neither the weak membrane model (U2 (f)) nor the Beta law Ui (f)) energies are convex functions. Consequently, we need to implement a global optimization technique to reach the solution. Simulated annealing (SA) cannot be used here because it leads to a prohibitive cost for calculations [9]. We have adopted a continuation method like the GNC [2]. 

Mephisto is devoted to predict the ultrasonic scans (A,B or C-scans) for a priori knowledge of the piece and the defects within. In the present version Mephisto only deals with homogeneous isotropic materials. The piece under test can be planar, cylindrical or have a more complex geometry. The defects can be either planar (one or several facets), or volumetric (spherical voids, side drilled holes, flat or round bottom holes). 

The basic requirements for the Mephisto model was satisfactory accuracy, that means prediction of amplitude, position and phase relation between the various signals, and short computation times, typically a few minutes for the simulation of a whole Cscan, compatible with an intensive use. These a priori contradictory characteristics have been contented by means of appropriate approximations based on physical considerations. 

In general, the assumption of quasi-stationarity is difficult to justify a priori. There may be several possible choices of intemiediates for which the assumption of quasi-stationary concentrations appears justified. It is possible to check for consistency, for example [A ]qj, [Pg.792]

Spatial synnnetry is one of the basic properties of a surface or interface. If the syimnetry of the surface is known a priori, then this knowledge may be used to simplify the fomi of the surface nonlinear susceptibility as discussed in section Bl,5,2,2. Conversely, in the absence of knowledge of the surface synnnetry, we may characterize the fonn of -iexperimentally and then make inferences about the synnnetry of the surface... 

An important point for all these studies is the possible variability of the single molecule or single particle studies. It is not possible, a priori, to exclude bad particles from the averaging procedure. It is clear, however, that high structural resolution can only be obtained from a very homogeneous ensemble. Various classification and analysis schemes are used to extract such homogeneous data, even from sets of mixed states [69]. In general, a typical resolution of the order of 1-3 mn is obtained today. 

The often-cited Amontons law [101. 102] describes friction in tenns of a friction coefiBcient, which is, a priori, a material constant, independent of contact area or dynamic parameters, such as sliding velocity, temperature or load. We know today that all of these parameters can have a significant influence on the magnitude of the measured friction force, especially in thin-film and boundary-lubricated systems. 

Wlrile tire Bms fonnula can be used to locate tire spectral position of tire excitonic state, tliere is no equivalent a priori description of the spectral widtli of tliis state. These bandwidtlis have been attributed to a combination of effects, including inlromogeneous broadening arising from size dispersion, optical dephasing from exciton-surface and exciton-phonon scattering, and fast lifetimes resulting from surface localization 1167, 168, 170, 1711. Due to tire complex nature of tliese line shapes, tliere have been few quantitative calculations of absorjDtion spectra. This situation is in contrast witli tliat of metal nanoparticles, where a more quantitative level of prediction is possible. 

For states of different symmetry, to first order the terms AW and W[2 are independent. When they both go to zero, there is a conical intersection. To connect this to Section III.C, take Qq to be at the conical intersection. The gradient difference vector in Eq. f75) is then a linear combination of the symmetric modes, while the non-adiabatic coupling vector inEq. (76) is a linear combination of the appropriate nonsymmetric modes. States of the same symmetry may also foiiti a conical intersection. In this case it is, however, not possible to say a priori which modes are responsible for the coupling. All totally symmetric modes may couple on- or off-diagonal, and the magnitudes of the coupling determine the topology. 

---