Roughness models for sanded wood surfaces

The understanding of the effects of variables is crucial to achieve the desired sanded surface quality at optimum condition. In wood surface evaluation, it is known that anatomies on wood surface could distort the roughness value and cause a misinterpretation of the processing performance. In this study, statistical approaches were taken to characterize the influence of sanding variables as well as to analyze the anatomical noises that were inherited from intra- and inter-species of woods. Four available roughness parameters (R _a , R _q , R _k and R _ap) were used to examine the surface of three distinct wood species, viz. kembang semangkok, red oak and spruce in wide-belt sanding. Based on the mean values, analysis of variance showed that species (anatomy) was significant to all conventional parameters except R _ap which was filtered by monitoring the second derivative of Abbott-curve. In spite of this, R _ap recorded a more widely dispersed deviation of random measurement values than R _k and R _a . The effects of grit size and feed rate were found to be significant. Empirical roughness models were established using response surface methodology, and the errors were calculated by comparing the model values to all the randomly measured values. Although exhibiting slight species-dependant effect by nature, R _k showed reliable consistency by recording the lowest error values (<10%) for both intra- and inter-species measurements. Experimental results also suggested that three random measurements at each run could be sufficient. The method of constructing machinability models can be readily applied in the industry as a quality control tool for wide-belt sander.