Sampling for the Measurement of Structured Surfaces

Structured surfaces are increasingly popular in manufacturing due to their ability to affect the function of a component, for example paintability and adhesiveness. Structured surfaces usually have complex geometrical structures on the micrometre to nanometre scale. These complex surface structures are challenging in terms of their measurement. For example, one widely recognised challenge comes from the increasingly high requirement of both measuring efficiency and measuring accuracy. Intelligent sampling is regarded as part of the solution to this challenge.

In this research, statistical sampling and signal sampling are investigated for the measurement of structured surfaces. Firstly, the widely used technique of uniform sampling is reviewed. Determination criteria for the sampling conditions of uniform sampling, i.e. the sampling intervals and lengths, are discussed. Four types of efficient (intelligent) sampling techniques, which were initially developed for the fields of statistics, computer graphics and coordinate metrology, are investigated. The intelligent techniques include: jittered uniform sampling, low-discrepancy sampling, model-based sampling and adaptive sampling. However, there are issues when applying these techniques to practical instruments; for example, they do not consider the measuring principles, such as the sensing mode or scanning method. Considering the measurement of surface topography, a sequential profiling adaptive sampling technique is proposed for raster scan-based stylus profilometers. Numerical evidence shows that the adaptive technique is promising for the measurement of linear patterns and tessellated structured surfaces.

To examine the performance of these intelligent sampling techniques, reconstruction techniques and error evaluation approaches are studied. A boundary segmentation algorithm has been developed to characterise the feature boundaries of surface features. With a sampling test toolbox, developed as part of the project, a sampling performance test is carried out in which the performance of seven selected sampling techniques is analysed. The experimental results show that adaptive sampling and model-based sampling methods have significant advantages over other methods. The proposed sequential profiling adaptive sampling has good performance in the measurement of linear patterned surfaces.

However, there are difficulties in fully enabling intelligence sampling for practical measurements. For example, the relationship between sampling and reconstruction has not been clearly understood. If the difficulties can be successfully addressed, intelligent sampling can be of promise in the next generation of measurement techniques.