A.B. Ivanov, A.S. Levchenko, A.F. Tashoyan, S.I. Kholshin

Investigating different physical media and processes occurring there we are inevitably facing the problems of effectiveness of our measurements analysis, i.e. selection of different regions or events on graphs (represented as phonograms, spectrograms, reflectograms, seismograms, cardiograms etc) of a specific form with subsequent determination of their numerical parameters of interest. Taking into account that these events are defined by the nature of investigated physical processes, it is necessary to use certain methods of determination of inadmissible deviations there, including a-priori specified masks and thresholds. Methods of identification of the events must ensure their reliable recognition - identifying forms and localizations, and the analysis - identifying the numerical parameters disregarding the scale of events and any possible distortions and noises. One of the solutions of this problem is to use the COMPAS technology (Complex Object-oriented Modeling, Prediction and Analysis of Similarity), which is a program combining the set of functions describing the events in details, setting the criteria of their determination as well as masks and thresholds on them. This program, apart from the standard data about events characteristic, can give an analytic function describing every particular event and the measurement error of its parameters. For the determination of regions which may contain one of the events, prescribed by its proper function, the whole data massive is processed using the wavelet-transformation. Conformity of wavelet-coefficients obtained to their coordinates allows us to find the region (a group of points) which contains the non-linear features (points of singular behavior of the experimental data), or in other words to determine the beginning and the end of each event. Identification of discovered (by means of wavelet analysis) features of the functional dependencies is realized by means of approximation of their localization regions with the special analytic function - model function, which provides the minimum discrepancy with the experimental data. Choose of such a function is realized by means of the functional minimization packet MUNUIT (minimizing chi square routine). After the events, i.e. the coordinates massive with the corresponding values are found and identified (determined like object-events), they can be completely excluded from the further analysis. That allows us to go from the experimental dependence to the object-represented dependence, which in turn allows the creation of mask-objects, definition of the limits of admissible experimental values as well as tracing the parameters of object-events. For each parameter, apart from its value the error is also defined which shows the allowed limits. Transmission to object-events is the problem of functional dependencies analysis - their division to distinctive fragments with their proper types and features. Obviously the COMPAS technology can solve the inverse problem as well (modeling problem), which comprises linking objects to known properties sets, from those one can get the functional dependencies describing investigated physical media and undergoing processes.