ABSTRACT
Manufacturers are always exerting significant effort to improve the quality of machined
parts by suitable choice of process parameters. Furthermore, there is a trend within
industry to improve process performance and product quality through analyzing
available historical data especially in chemical industry. This trend is driven by the need
to reduce product development time and cost. The use of latent variable modeling using
historical data has been proposed in the past for product design and quality
improvement (C.M. Jaeckle and J.F. MacGregor) [23]. This paper outlines the
application of such approach using Projection to Latent Structure (PLS) and its model
inversion to facilitate the choice of cutting parameters for a desired surface roughness
while maximizing the Metal Removal Rate (MRR). The approach is mainly based on
using historical data readily available on most of factory platform and simulated through
experiments conducted on three different milling machines under normal conditions
(sharp tool and stable cut). The model inversion approach is formulated in an
optimization problem using the latent space linear model with nonlinear constraint. The
approach output solutions were validated with the results showing that the proposed
technique can be used for process planning and quality improvement of machining
data.