Development of a Hybrid PCA-ANFIS Measurement System for Monitoring Product Quality in the Coating Industry

In industr?, todq many producfs are soldfortheir efficacy rather than their chemical composition. Thereare several key attributes within the coating industrv suchas, Anchorage, Seal strength etc., which characterize thequality of the final product and are features used by thecompany to promote the sale of the product. Such qualifyvariables (dependent variables) however may i.ivolvemeasurement difficulties. The df$culties can be due to avariety of reasons, including: (I) Reliabilify of on-linesensors. (2) Lack of appropriate on-line instrumentation. Inthe coating process off-line laboratory tests dett:nnineproduct qualify measurements. However, such laboratoryanalyses introduce dela-vs in the measurement if keyperjormance indicators. This can result in a signz@canteconomic loss’ if the analysed product fails the qualityconhol test. An improved monitoring system is requiredtherefore to determine product quality online and miviniisecommercial wastage. To facilitate this, advancedmoniroring and control or optimization techniques requireinferred measurements, generated with correlations ?omreadily available process variables (indep,mdentvariables). Although inferential models are widely used inindustr?,, only a few techniques for inferential modeldevelopment are discussed in the open literature. Thispaper therefore will present an improved systematicapproach for the development of inferential models usingsoft computing systems and demonstrate the methodologyby inferring the ‘Anchorage ‘ of polvmeric-coatedsubstrates (i.e. Tpek orpaper) in the coating industv.