DD205998A1 - METHOD AND ARRANGEMENT FOR THE TECHNICAL DIAGNOSTICS OF MACHINES - Google Patents

Abstract

THE INVENTION CONCERNS OPERATIONAL MONITORING AND DAMAGE TO ROTATING MACHINES. THE OBJECTIVE AND TASK IS TO APPEAR AN ALGORITHM FOR REAL-TIME SIGNAL PROCESSING IN ROTATING MACHINES, WHICH ALLOWS MINIMIZED EFFORT AND HIGH RELIABILITY TO SPECIFY DERIVED SIZES WHICH, ESSENTIALLY, ARE CHANGED ONLY BY DAMAGE PROCESSES TO CREATE EXACT CONDITIONS FOR CLASSIFYING SHAMS. THIS IS ACHIEVED THAT MULTIPLICATIVE OR ADDITIVE CORRECTION FUNCTIONS, IN PARTICULAR, ELIMINATE FROM THE MIDDLE AMOUNT OF THE STOCHASTIC SIGNAL ALLOCATIONS OF THE VIBRATION SIGNALS MEASURED ON THE MACHINE BY THE OPERATING AND ENVIRONMENTAL PARAMETERS IMPACT. THE CORRECTION FUNCTIONS ARE DIFFERENTIAL EQUATIONS, THE COEFFICIENT OF WHICH IS VARIATED BY A SEARCH ENGINEMETRY PROGRESSIVELY IN INTAKE MACHINE, UNLESS AGAINST CORRIGATED SIZE AND THE OPERATING AND ENVIRONMENTAL PARAMETERS, NO STATISTICAL DEPENDENCE CAN BE DETERMINED. THE FIELD OF THE INVENTION IS THE MEASUREMENT TECHNOLOGY.

Description

24 102 1 1

Title of the invention

Method and arrangement for technical diagnostics of machines

Field of application of the invention

The invention relates to the field of technical diagnostics of machines, in particular of power plant machines, such as turbines and pumps.

Characteristic of the known technical solutions

For operation monitoring and technical diagnostics of machines are preferably vibration signals, such. 8. Lagerbockschwingungen and wave vibrations and temperatures such. As storage and oil temperatures, measured and registered. The evaluation is carried out by determining limit value overruns in the frequency domain and in the time domain. Corresponding methods and arrangements are shown in the patent specifications DE-OS 215 2848 and DD-WP 105 068 and in the company documents of Carl Schenk AG, Darmstadt: Vibrocontrol 2000 and Vibrologger. Methods and apparatus for forming center functions, spectra obtained by FFT and digital filters have been disclosed in Brüel & Kjaer Technical Review 2-1975, 3-1975, 1-1978, 1-1979, 2-1980. In the operating documentation of the stochastic NSA 1000 analyzer from EMG Budapest, the formation of mean functions of auto and cross-correlation functions is described procedurally and in terms of equipment.

241021 1

Furthermore, an evaluation of the signals by a statistical signal processing and feature formation for characteristic damage using methods of pattern recognition by D. Barschdorf, Technical Diagnosis of Machines with rotating parts, in Scientific Reports IHS Zittau, IHZ- & K-335-79 has been shown. Similarly, ΏΏ-ΉΡ 134 991 proposes a method for the separation of periodic and stochastic signal components.

The common disadvantage of these known methods for technical diagnostics is their low reliability. Methods for the reliable detection of damage from measurement results are so far unknown.

Object of the invention

The object of the invention is the development of a reliable method and arrangement for the early detection of damage from signals measured on rotating machines _; and their effort is minimized.

Explanation of the essence of the invention

The invention has for its object to provide an algorithm for the real-time signal processing of signals measured on rotating machines and comparison variables for the formation of derived quantities that change essentially only by damage processes and the conditions for the classification of damage to create «The object of the invention is achieved by eliminating the influence due to operating parameters and environmental parameters by means of multiplicative or additive correction functions from the functions or the mean values of the signals or from variables derived therefrom, in particular from the mean amount of the stochastic signal component of oscillation signals or from spectral amplitudes · The correction functions are, in particular, difference equations whose coefficients according to the invention are based on a search algorithm

24102 1 1

be varied stepwise with an intact machine until no more statistical dependence between corrected size or function and the operating and environmental parameters can be determined. This is determined by approximately zeroing the cross-correlation functions between operational and environmental parameters and corrected magnitude. After such a coefficient adjustment, it is possible to unambiguously make a statement about a machine damage by exceeding the limit value of the corrected quantities.

The method described can be realized by analog, digital and hybrid computers and by an interconnection of conventional measuring devices with special circuits or even microcomputers

 The arrangements according to the invention for carrying out this method are characterized in that they are hardwired or programmed by algorithms for implementing a correction function, in particular by means of a digital filter, algorithms for subtracting the output of the correction function and the input variable to be corrected, a cross-correction algorithm for the corrected size and Operational and / or environmental parameters, a switch network or an algorithm for selecting a respective Betrisbs- or Umweltoarameters as an input of the correlator and an algorithm for gradually changing the coefficients of the correction algorithm included.

Furthermore, they have a classifier that detects and classifies damage from the sets of coefficients of the correction function obtained with statistical independence.

Working Example

The invention will be described with reference to an embodiment by means of Fig. 1 and Fig. 2 in more detail. Show it

24102

Fig. 1 shows the sampled time functions for a current time signal X and a center function or "the quasi-periodic signal Y ^ a measured vibration during the η-th rotor rotation of the machine

2 shows the arrangement according to the invention for early damage detection on rotating machines

The inventive method for technical diagnostics on machines provides that from the current vibration signal X, which is measured on the rotating machine, by preprocessing the average amount of the stochastic signal component S according to the equation

I - ^S n-1

where 8-. X »- Y". »

and Y = ¹ . (X ¹¹¹ - Y ^m .) A, + Y ³¹ ,

π ^v η n-1 ' 2 n-1

with the averaging coefficients a ,, a. <ir 1 is to win.

the current amount of the mth IVe

Dabei.bedeuten

t 'of the stochastic signal component during the η-th rotor revolution of the engine, X' the mth sample of the current vibration signal X during the ηth rotor revolution and Y 'the mth value of the center function of the vibration signal during the ηth rotor revolution. Parallel to this, the actual measured quantities P., P ₂ ,... Of the operating and environmental parameters of the rotating machine supply the correction function or quantity. This correction function or variable and the mean amount of the stochastic signal component S are subtracted from the corrected function or magnitude. Furthermore, by stepwise modification by an optimization algorithm, the coefficient matrices A, .3 of the correction algorithm are to be set such that the cross correlation function of corrected function or magnitude and current measured quantities P ₁ , P ₁ ... Is approximately zero.

y f 4 3 i ϊ '

By a subsequent limit value monitoring and trend determination via the autocorrelation function of the corrected function or variable S, the damage prediction or recognition takes place. In addition, the classification of the damage is carried out by analyzing the coefficient matrices A, 3 with the aid of damage patterns SM with the aid of pattern recognition methods.

The arrangement according to the invention for carrying out the method according to FIG. 2 provides that the operating and environmental parameters P 1, P 2,... Measured for the rotating machine 1 to be monitored for realizing the correction function of the correction circuit 2, e.g. B. a digital filter whose output is coupled to the positive input of the subsequent subtractor supplied. At the minus input of the subtracter 3 is the average amount of the stochastic signal component S, which is obtained in the upstream preprocessing unit 4 from the measured signal of the rotating machine. The output of the subtracter 3 provides the corrected function or magnitude S i of the signal and is connected to the input of the subsequent threshold comparison circuit 5 as well as to the X input of the correlator 5. The Y input of the correlator 6, the operating and environmental parameters P., P ", ... of the monitored machine 1 are also fed via a switch network 7. The cross correlation signals KKF generated in the correlator S are supplied from the output thereof to the input of the optimization circuit 8 for the stepwise change of the coefficients A., B .. and for the detection of the statistical independence of the corrected variable or function from the operating parameters P *, P ~, .. . guided.

On the one hand, this optimization circuit 8 controls the switch network 7 and, on the other hand, the coefficient outputs A _{i #} B ^ are coupled to the coefficient inputs of the correction circuit 2 and to the one input of the classification circuit 9.

241021 1

The second input of the classification circuit 9 is supplied with the coefficient sets of damage examples SM. This arrangement according to the invention is characterized in that, independently of the operating conditions of the machine 1 to be monitored, the normal and abnormal behavior of the machine can be reliably recognized. Furthermore, the invention allows an exact trend prediction and early damage detection for rotating machines for the first time. The correction circuit 2, the subtractor 3, the correlator 6 and the optimization circuit can preferably be realized by microcomputers.

Claims (6)

24102 1 1 Erfindunqsanspruch 1 · Method for the technical diagnostics of machines, in which in a known manner mean functions, averages, spectral amplitudes, average noise squares, average amounts of the stochastic signal component and cross-correlation functions are formed, characterized in that from the time function, from the averaged time function or from the spectral amplitude of a vibration signal or from a temperature signal or from the mean square or the mean amount of the stochastic signal component of vibration signals, which are measured on the machine, which is eliminated from operating parameter changes and environmental influences of the machine portion, by means of Cross-correlation measurements between corrected function or corrected value and operating parameters or other environmental variables, the coefficients of a multiplicative correction function or a correction factor or an additive correction function or a correct urgröße particular a time-dependent difference equation are set so that the corrected function or the corrected size of the operating parameters and environmental variables is independent. 2. A method for the technical diagnostics of machines according to item 1, characterized in that the coefficients of the correction function with intact machine and the emergence of damage by Trendermittiung also by means of the autocorrelation function of the corrected size or function and. to be determined by limit value comparison. 3. A method for the technical diagnosis of machines according to item 1, characterized in that the Trenderraittlung and Grenzwertvergieiche to the coefficients of the correction function takes place, which are constantly adapted to the changing properties of the machine and thereby the formation of damage to the machine is fixed. 241021 1 4. A method for the technical diagnosis of machines according to item 1, 2 and 3, characterized in that on the basis of observations of damage coefficient classes are formed, which serve with the aid of pattern recognition methods for the classification of damage. 5. Arrangement for carrying out the method according to item 1 and 2, characterized in that with the plus input of the corrected magnitude or function (S ,,) supplying subtractor (3) to be corrected size or function (S) and the off -IQ gear of a correction circuit (2), in particular a digital filter, is connected to the signal inputs the operating and environmental parameters (P ₁ / ^p ₂ '···) ^ ^he machine ^{to be} monitored (1) are turned on and their coefficient inputs with the coefficient outputs ( A, B) are connected to an optimization circuit (8) for the stepwise change of the coefficients, that the minus input of the subtractor via a preprocessing unit (4), which provides the average Setrag the stochastic signal component (5_) from the signal measured on the machine Coupled with the machine is that the input of the optimization circuit leads to the output of a corrector (S) whose X input is connected to the output of the subtrah ierers and its Y input via a controlled by the optimization circuit switch network (7) are connected to the recording device of the operating and environmental parameters of the machine and that at one input of a classification circuit (9) the coefficient outputs (Α., 3_. ) of the optimization circuit and to the other input of the classification circuit the coefficient sets of damage examples (SM) are connected. 6. Arrangement for carrying out the method according to item 1 to 5, characterized in that the correction circuit, the subtractor, the correlator and the optination circuit for stepwise change of the coefficients by digital computer, in particular microcomputer, are realized. To this 1 page drawings