Design and Development of an Efficient Algorithm and Its Software Code for Performance Analysis of Parallel Multi -compressor System

This paper analyses a multi-compressor system for its performance failures and subsequent improvements. The logbook data of this system has been obtained. Data has been classified using various state-of-art data classification techniques. This paper presents a comparative analysis of Fuzzy clustering algorithm, Hard-c-means clustering and Gustafson-Kessel clustering algorithm. Data clustering errors have been obtained and results have been plotted. Concept of Fuzzy entropy as a measure of error has been implemented in this paper for the first time. Finally based on this data handling, data based controllers for multi-compressor systems have been developed using back-propagation algorithm and bi-directional associate memory neural networks. It has been shown that the controller error is less in the case of backpropagation algorithm. Thus, finally authors recommend fuzzy clustering for data handling and backpropagation algorithm for controller development for this industrial system.