Analysis of Noise Emitted From Electrical Machines by Clustering Method

Abstract

This paper refers to a study of noise, emitted from different electrical machines (AC & DC) as noise is very noticeable and physically harmful to human beings. This study has been conducted in an electrical machines laboratory. Different noise parameters such as LAeq (equivalent continuous A-weighted sound level), LAE (sound exposure level), LAV (average sound level) and TWA (Time weighted average level) are taken by a precision noise dosimeter (Model-4444). Clustering method is applied for analyzing high and low noise variables for identical noise producing machines. Clustering is a popular non-directed learning data mining technique for partitioning a dataset into a set of clusters (i.e. segmentation). Although many clustering algorithms may be applied, but none of them may be superior on all datasets, and so it is never clear which algorithm and which parameter settings are the most appropriate for a given dataset. Our dataset is based on the different sound parameters mentioned above. We have applied graph based clustering method (Zahn’s algorithm) for analyzing this. Zahn’s algorithm first constructs a minimal spanning tree and then deletes the inconsistent edges i.e. the edges which are too long as compared to the other edges, to form clusters. For the construction of the minimal spanning tree we have used the Prim’s algorithm. The result implies grouping of similar noise producing machines. In the first iteration there were two clusters. The second iteration resulted to three clusters. We have shown up to third iteration which resulted to the formation of four clusters of machines producing four identical levels of noise.