Study of the improved algorithm for correlation integral method and its application to surge precursor identification.

Correlation integral method (CIM) is a perfect method for the detection of stall and surge. However, CIM is very time-consuming due to its computational complexity. This paper aims to resolve the disadvantage. First, the segmented computation of correlation integral (CI) is deduced; second, the mapping criterion is given; finally the improved correlation integral method (ICIM) is proposed. The criterion can search the repeated subset of points and pair of subsets efficiently, which aids the ICIM in eliminating all redundancy caused by the overlapping of consecutive time windows. By applying ICIM to the detection of surge precursor in centrifugal compressors, it is illustrated that CI has a sharp decline more than 1.5 s prior to the fully developed surge. Also the processing time of ICIM is much shorter than that of CIM. The explicit detection of surge precursor, its rapid recognition, and the resulting longer time available for the feedback preventing the surge are the principal advantages of this approach. These studies are significant for CIM and will boost its application in numerous fields, especially in the detection of stall and surge. [ABSTRACT FROM AUTHOR]