Variability analysis using hybrid intelligent methods and sequential Monte Carlo simulation: application to quality control of material

Abstract

This paper is concerned with a method for quality control– and evaluation-based measurements and advanced methods of modeling and identification systems. The variability of properties is measured by the magnitude of the distance between the output of the model and the required measurement; the variability level depends on the several factors affecting the distribution of the measurements during different manufacturing and operational steps. The objective of this work is to evaluate the variability of the properties of material using inferential model and measurements; it usually used multivariate regression. The main local chemical elements of material at a specified point obtained by a scanning electron microscope (SEM) are analyzed using a hybrid scheme based on the identified measurement models and Monte Carlo simulation. The variability assessment is quantified by modeling errors obtained by the propagation of random input changes through the model. A comparative study between different techniques based on intelligent methods is given; the variability is monitored, and control limits are defined using statistical tools. This kind of approach is applied for constructing an accurate computing procedure for variability analysis and quality assurance.