Pattern recognition in multivariate statistical process control for dimensional transformation of statistical parameters

In this paper is proposed a method of multivariate pattern recognition applied in a industrial case of statistical process control with two quality characteristics. The essence of functioning of the method is the suitable determination of the values of the weights of the activation function (FA) of the artificial neural network (RNA) perceptron, so that without using the algorithms of this training set is used dimensional transformer as an asset. The application of the FA of the perceptron with the approach used in this research, is novel and an alternative to the common use of the same. The industrially proven method proved to be of relevance in the multivariate statistical process control. The bivariate case was analyzed by combinations of patterns that compose the variables X1 and X2 that integrate the binary variable X = [X1, X2]. The experimental and industrially analyzed patterns for these variables were found in the real case study and are: natural, increasing trend, decreasing trend, upward shift and downward shift. The results show a high rate of recognition of patterns in the random variables that comprise each bivariate combination.