Goals of the laboratory. The laboratory is created to develop new approaches for processing big volumes of complex geophysical information using fuzzy logic and fuzzy mathematics methods. Development of new mathematical approaches and adaptation of existing methods of pattern recognition for studying the Earth's magnetic field and solving other geophysical problems are provided. One of the priority activities deals with development of methods for recognition of disturbances with defined morphology in geomagnetic data time series. Activities of the laboratory are implemented in the following directions:

1. Pattern recognition
   • Development of theoretical and algorithmic basis of recognition
   • Creation of authoring software products
   • Recognition tasks on magnetic data
   • Other geophysical and geological recognition tasks
2. Deployment of magnetic observations and creation of INTERMAGNET Russian segment
   • Creation of five new INTERMAGNET observatories in Russia as joint observatories of GC RAS and institutions of regional RAS branches
   • Implementing mobile magnetic observations
3. Creation of national geomagnetic data node at GC RAS servicing Russian INTERMAGNET observatories
   • Transmission of magnetograms from functioning Russian INTERMAGNET observatories to GC RAS in real-time mode
   • Publication of pattern recognition algorithmic software on the Web
   • Maintenance of storage of Russian observatory magnetograms in databases

Relevance of research areas. The present stage of geophysics development is characterized by transition from general models largely based on the use of general physical laws, to models describing specific geophysical objects and structures. Objects of geophysical investigation are likely to have complex internal structure and vary with time. For such objects criterion of recurrence of experiment conditions might not be satisfied. This results in a certain common character of methodological approaches between geology, geophysics and areas of humanities with considerable fuzziness of applied concepts and characteristics and a great role of hardly formalized opinion of individual experts.

Fuzzy information is usually processed by experts. However, due to a huge volume of initial data, such solution could be practically unrealizable. Thus, there is a problem of automatic analysis of huge volume of factual information with a lack of clearly and definitely formulated criteria and features of studied objects. To solve these problems methods of pattern recognition including expert systems and algorithms of data processing based on fuzzy logic are used. Besides, methods of clustering and morphological analysis are developed and applied for these purposes.

Comparison between results obtained with classical approaches (e.g., statistical methods) and fuzzy logic methods developed by GC RAS allows to assume that the latter turn to be more efficient in the case of a certain lack of factual information and/or in the case of processing more complex signals. In these cases classical methods give less stable and therefore robust results. On the contrary, classical statistical methods prove to be more efficient when there is a sufficient data volume for statistical processing (due to greater simplicity of realization and greater certainty of obtained results).

Major achievements of the laboratory. A new approach for working with discrete data entitled Discrete Mathematical Analysis (DMA) is created. It is based on modeling the discrete analogues of fundamental mathematical concepts using fuzzy logic techniques. DMA represents a series of algorithms for data analysis and processing: clustering, tracing, smoothing and prediction of time series, their morphological analysis, trend detection, etc.

DMA includes two groups of data analysis applications. The first group deals with recognition of higher density subsets in finite metric spaces. This includes RODIN clustering algorithm, CRYSTAL and MONOLITH algorithms for detection of higher density areas in multidimensional geophysical data arrays. The second group deals with recognition of disturbances of different nature in time series. DRAS (Difference Recognition Algorithm for Signals), FLARS (Fuzzy Logic Algorithm for Recognition of Signals) and FCARS (Fuzzy Comparison Algorithm for Recognition of Signals) algorithms created in the framework of DMA represent modeling of a data interpreter’s logic for its further automated application to large data array processing. Basing on DMA and fuzzy logic a series of specific algorithms for automated recognition of anomalies on magnetograms is being developed.

In collaboration with the Laboratory of Geophysical Data the regional geomagnetic data node of the Russian INTERMAGNET segment is being created on the basis of these two laboratories. A particular feature of this node is the automated system for recognition of artificial disturbances on incoming magnetograms, which is being introduced. INTERMAGNET network is the basis for geomagnetic field monitoring so requirements for reliability of collected data are very high. Therefore, an important task is an objective and formalized recognition and further elimination of possible technogenic anomalies in data records. Algorithms SP and SPs for spike detection and JM algorithm for recognition of baseline jumps are designed for processing data recorded with 1 minute, 1 second and less sampling rate.

The algorithmic system created with the use of fuzzy mathematics for geomagnetic records enables producing filtered, despiked magnetograms from preliminary data almost without human intervention.

Online 1-second magnetogram de-spiking
BETA VERSION

Input file format: IAGA-2002	The SPs algorithm is a tool applicable to any time series that have specific time anomalies (disturbances), which have to be identified. The algorithm is specifically aimed at recognition of singular artificial spikes with a simple morphology on 1-second magnetograms. An example of such spikes, generated by a nearby running truck and reflected on magnetogram, is given in the figure.
SPs free parameters:
1 channel nu: 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49 0,50 0,51 ro1: 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49 0,50 ro2: 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49
2 channel nu: 0.45 0.46 0.47 0.48 0.49 0.50 0.51 0.52 0.53 0.54 0.55 ro1: 0,36 0,37 0,38 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 ro2: 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49
3 channel nu: 0,51 0,52 0,53 0,54 0,55 0,56 0,57 0,58 0,59 0,60 0,61 ro1: 0,38 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 ro2: 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49
4 channel nu: 0.45 0.46 0.47 0.48 0.49 0.50 0.51 0.52 0.53 0.54 0.55 ro1: 0,38 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 ro2: 0,39 0,40 0,41 0,42 0,43 0,44 0,45 0,46 0,47 0,48 0,49
SPs algorithm description