Thesis for competition for a philosophy doctor in engineering degree on a specialty 05.11.01 - Devices and methods of measuring mechanical quantities, National Technical University of Ukraine "Kyiv Polytechnic Institute", Kyiv, 2008.

In this thesis, the present state of research in the field of diagnostics and condition monitoring of rotating machinery, and especially gas-pumping equipment, is broadly covered. On the basis of the analysis carried out by the author, it's concluded, that making automatic diagnostic systems is very prospective trend along with making intelligent diagnostic systems.

An object of the investigation is the new gas-turbine engine driven gas-compressor unit GPA-25/MN80.02, and in particular the 650-21-2 rotary charger rotor system. As the DN-80 engine that drives the charger, is under SCADA system control and diagnostic surveillance of the producer (NPO "Zoria-Mashproekt" plant), and the charger is left out of control, the decision is made to create a method for diagnostics of a charger along with predictive condition monitoring.

First of all, mathematical and generational models of most common 650-21-2 charger defects are formulated. Analytical formulae for forces and moments in journal bearing with misalignment are obtained. Symptoms for "misalignment in journal bearing" are obtained on their basis from the mathematical modeling of the charger rotor-bearing system. It should be noted that obtained formulae can be used to model a wide range of machinery defects, including bent shaft (i.e. thermal or mechanically induced bow), angular and parallel bearing axis misalignment, etc. The only boundary of usage for the formulae is eccentricity value more then 0.6…0.7. The diagnostic chart (defects along with their symptoms) for the charger is formed based on competitor's research results and data of used references. That chart allows detection of such defects as bearing fastening looseness, journal bearing asymmetry, misalignment in journal bearing, parallel and angular misalignment of drive and charger shafts and charger shaft bow.

One of the vital tasks for diagnostics of rotating machinery is the determination of machine's natural frequencies. However, this task is rather sophisticated for automation, as no method allows one to fully formalize and automate the determination. A new algorithm for machine's natural frequencies determination is proposed, thus providing means to automate this sophisticated procedure. This algorithm combines two methods - peak amplitude method and inverted APFC method to obtain correct results, and allows making the natural frequency detection semiautomatic. The algorithm has beet successfully tested on dynamic model of the charger shaft. The proposed algorithm should not be used for the machinery with closely spaced modes; for such machinery the careful analysis of APFC and inverted APFC is necessary (peak-amplitude method fails in most cases of analysis of closely spaced modes).

One of the tasks of predictive condition monitoring is a prediction of the technical condition for a given machine. However, it's possible that due to transducer faults or short circuits one obtains abnormal data, that doesn't reflect current condition of the machine. There are many abnormal data elimination methods described; however, almost all of them can process only periodic (or almost periodic) data or (statistic-based methods) require large samples of data (at least 50-100 points). An algorithm for abnormal data elimination, which can process non-periodic and small samples of data, and allows easing of the data analysis for machinery state prediction significantly, is also proposed. This algorithm is included in developed condition monitoring and diagnostic system of gas-pumping station equipment software and has been successfully tested on live data.

Most of current stationary monitoring and diagnostic systems provide information about shaft orbit. However, the analysis of this kind of data is rather complicated, as, for example, one cannot determine the frequency composition of the orbit. A new vibration data analysis method that solves this problem - decomposition of shaft orbit into Lissagou figures (Ukraine patent UA 28204) is proposed. This method allows one to determine the frequency composition of the orbit (both in frequency and order domain) and a direction of precession for this frequency. The result of the decomposition is a set of vibration ellipses along with the precession directions, which can be used both for condition monitoring and diagnostics.

Three methods of establishing statistically founded alarm limits for decomposition data, that are to be used in the condition monitoring of machinery, are proposed. These methods are based on the assumption, that vibration has two-dimensional normal (Gaussian) distribution, and allow one to establish profiles of different forms, based on rectangular-shaped, elliptic-shaped and scatter-ellipse confidence regions. Each created profile corresponds to machinery technical condition with a given confidence probability.

On the basis of the above research, two methods of 650-21-2 charger diagnostics (using and not using decomposition of shaft orbit into Lissagou figures) are developed. Those methods are intended to be used with the created condition monitoring and diagnostic system.

After all, on the basis of selected ideology and carried out criticality analysis of gas-pumping station equipment, a condition monitoring and diagnostics system for gas-compressor units GPA-25/MN80.02 is created. This system meets the requirements of current Ukrainian national and international ISO standards in the field of condition monitoring, diagnostics, vibration measurement and so on. The system first started in 2004 and put into operation in 2007. The results of the field tests and operation have proven that the system in question successfully solves condition monitoring and diagnostics tasks.

Keywords: vibration, technical diagnostics, condition monitoring, gas-compressor unit, rotary charger, journal bearing.

Размещено на Allbest.ru