Abstract
In implementing multivariate statistical process monitoring (MSPM), a frequently encountered problem is the estimation of Confidence Bound, which in fact, is a problem of density function estimation. Traditional assumption of normal distribution of process data often mismatches the real situation. As a new non-parametric method for density estimation, neural network estimator is proposed to estimate the distribution and density function of multivariate statistic. Estimated result is used to calculate the Confidence Bound of multivariate statistical process control (MSPC) chart. Experiment study illustrates the proposed technique has the permit of simplicity and effectiveness.
