Development of a system for diagnosis of the Presurge State of a centrifugal compressor

Abstract

Surging is an unacceptable event observed during operation of a centrifugal compressor (CC). The high probability of damage to the rotor, bearings, and seals during surging makes it necessary to utilize antisurge protection systems. The basic principle of the design of such systems is related to measurement of the output corresponding to the boundary of the surging event. Because of the multiparameter dependence of output on variable factors (gas intake parameters, rotation rate, etc.), a socalled "setting" is introduced, i.e., the increased output value at which the antisurge protection system will switch on. This limits the range of acceptable operation of the compressor. Another drawback of this method of designing the antisurge protection system is the necessity of precise measurement of the compressor output, which is somewhat complicated in many cases (and even impracticable, for example, for natural gas compressors). The proposed system of diagnosing the presurge state of CC is based on an analysis of the signs of nonstationary processes that precede the onset of surging. The experiments conducted [ 1] show that in the overwhelming majority of cases a rotating stall precedes surging in the casing of CC. This process usually arises near the surge boundary, and the antisurge protection system should be switched on when it appears. The design concepts of the proposed system were experimentally investigated in the Department of Compressor Design at St. Petersburg State Technical University. The experiments were conducted on a single-stage model of standard and promising CC with an impeller + diffuser + rotary elbow + reverse control apparatus scheme. Impellers (Im) of the closed type of different geometry combined with vaned (VD) and vaneless diffusers (VLD) with peripheral velocities of up to 280 rn/sec. A measurement complex developed at the SPSTU was used; it consisted of a miniature static pressure pulsation sensor installed on the wall of the diffuser; communication line; eight parallel channels with a memory volume of 68 kilowords per channel; a synchronization system with tracking of the measurement channels; a system for measuring the rotation system; and a minicomputer. Signals from a sensor positioned on the midline of the intervane channel at the end of the diffuser are given as an exampie of processing. The signals were analyzed using the correlation theory of random processes and methods of spectral analysis [2]. To exclude excess information, filtration with a band filter with a low cut-off frequency of 3 Hz and a high cut-off frequency of 100 Hz was conducted. The correlation functions were calculated with the equation