Visualisation and dimension reduction of high-dimensional data for damage detection

Important developments have occurred recently in the field of damage identification as a result of the import of numerous techniques from the disciplines of multivariate statistics and pattern recognition. One problem in the application of these methods is the curse of dimensionality, which can complicate and sometimes invalidate the use of certain techniques if the data under examination has too high a dimension. The object of this paper is to illustrate the use of some established methods of data visualisation and dimension reduction on some data sets of Engineering interest. The methods of reduction include: simple projections, linear and nonlinear principal component analysis and Sammon mapping. The paper also investigates the use of one-outlier detecting components as a means of dimension reduction. The data sets include examples from simulation and experiment and the application is to machine and structural health monitoring.