The Objective Selection of Multi-Dimensional Technical State Characteristics Vectors

In the paper, the method based on Sammon's transformation is presented, enabling the visualization of multi-dimensional data in a two-dimensional plane - being friendly for human perception. This transformation is executed in the form of the implementation that makes use of self-organizing learning process. The evaluation of separation of the initial vectors and vectors on the plane is based on the implementation of modified Sebestyen's criterion. The evaluation considers the distances between classes and internal dispersions within the classes. Visual, two-dimensional Sammon's representation of state vectors and the application of Sebestyen's criterion provide rational selection of symptoms essential for the classification of technical condition of machines and devices.