A.M. Aliyev FUZZY ALGORITHM FOR PLANNING AND CONTROL EXPERIMENTS IN GAZLIFT WELLS

S.F. Jafarov ELECTROMAGNETIC VIBRATION EXCITER’S MECHANICAL PARAMETERS SPECTRAL ANALYSIS ALGORITHM FOR FUZZY CONTROL

The problems of electromagnetic vibration exciter’s (EVE) output parameters spectral analysis are considered. This EVE represents an electromechanical converter, which consists of set the different units distinguishing by the masses, stiffness, etc. Therefore when we analyze the processes of EVE, a number of the harmonics, their amplitudes and phases, also a number of the estimated spectrums’ ordinates are very big.

One of the opportunities to calculate the estimates of a spectral density during the assumed time under the N-dimensional array of an initial information, is an application of an algorithm of fast Fourier transform. At first, due to this algorithm we calculate the complex Fourier coefficient of a vector random process in the section electromagnetic vibration exciter- technology process.

We combine all ordinates of the estimated spectrums in the groups with а number of N and a width of  (frequency step) and as an estimate of own (reciprocal) spectrum we use an arithmetical average of the ordinates in band . In this case we receive two positive effects:

The dimensions of a control algorithm decreasing;

Due to averaging of the spectrums estimates in the frequency bands, we receive supplementary decreasing of their dispersions and the different in a shift of the estimates, which are received by smoothing by the different windows become insignificant.

Thus, as the estimates of the own and reciprocal spectrums we use N values of: 1) the own spectrums smoothed estimates 2) the absolute values and phases of the reciprocal spectrums, which are averaged in the frequency, band. These estimates are compared with the lues of the spectral matrix’s elements at the same frequency range. As a result we receive R-dimensional (R=Nn, where n- a number of the estimated spectrums), vector of the mistakes (). A vector of the mistakes is an initial information for an iterative control algorithm, that is a kernel of a system making decision. This way lets system make more flexible decisions at the fussy control of the vibration parameters.

S.V. Zhernakov HETEROGENEOUS KNOWLEDGE BASE FOR GAS TURBINE ENGINE PARAMETER DIAGNOSIS AND CONTROL

An interaction of heterogeneous (different) knowledge bases (database (DB), expert (EKB) and conceptual (CKB) knowledge bases, rule base (RB)) with a gas turbine engine mathematical model (MM) and external utilities during diagnostic and control process is shown on Fig.1. The general procedure of knowledge base (KB) request could be represented as the following:

,
is a component object of KB; is a concept of KB; is a set of conceptual model input and output parameters.



Fig.1 Heterogeneous KB structure in hybrid expert system environment

In order to use several KB and DB in expert system, these bases must be united in unified information space by input programming interpreting language which allows to effectively adapt and expand the conceptual model of gas turbine engine.

The solver of expert system performs decision of the certain problem by the following request:

MM(S1)  MM(S2){t};  ?S1,S2

MM(S1) is a reference mathematical model of gas turbine engine under test;

MM(S2) is a mathematical model of defective gas turbine engine;

S1, S2 are parameters to be under diagnosis; {t} is a logic operator.

EKB is represented by exact, fuzzy and combined production (“IF X, THEN Y”):

X is a precondition represented by D1  D2  ... Dm, where Di is a disjunction represented by E11&E12&...E1n; Y is a inference represented by G1&G2&...Gk. A fuzziness in the EKB can be represented as: (X  Y(Z)), where {X},{Y},{Z} are fuzzy sets defined in universal sets {U},{V},{W}. The X,Y,Z sets partially or completely define indefinite parameters of gas turbine engine fuzzy diagnostic model.

The gas turbine engine diagnostics in expert system EKB can be represented as the following: Task(X)  Subtask 1(X1)#...#Subtask N(Xn), where X is a set of task input and output parameters.; # is a symbol of pass from one EKB to another.

So, every task (target) can be split up into several subtasks (subtargets). In this case, the selection of appropriate method to solve a basic task depends on specialties of each subtask solving. The parameter controls the objective task solving is a specialty of its solving process. Therefore, the production of EKB task is transformed in subtask’s productions: Task(X) = Specialty  Subtask1(X1) = Specialty1#...#Subtask N(Xn)=Specialty n.

The proposed approach to heterogeneous knowledge base development for gas turbine engine diagnosis and control based on static and dynamic hybrid expert systems allows the following:

- to facilitate the adaptation of the object under diagnostic;

- to apply different knowledge sets (rules and fuzzy rules in database, knowledge base, expert knowledge base) including different inference algorithm for efficient diagnostic problem solving;

- to create a powerful hybrid information system with a simultaneous graphic interpretation of object under test;

- to include software units for sensor imitation.