Process Monitoring: an Alternative Approach for D2 Contribution Plots

CARLOS R. ALVAREZ; ADRIANA BRANDOLIN; MABEL SÁNCHEZ

Buenos Aires, Argentina

Congreso; XXII Interamerican Congreso in Chemical Engineering y V Congreso Argentino de Ingeniería Química (ISSN 1850 3519, ICChE – ISSN 1850 3535, ACChE); 2006

AAIQ (Asociación Argentina de Ingenieros Químicos) - IAACHE (Interamerican Confederation of Chemical Engineering)

En Libro de Proceedings publicado en CD (Nro. págs: 15) Abstract. The Hotelling statistic (D2) is widely used in Statistical Process Control to reliably detect the out of control status, but offers no assistance as fault identification tool. A solution to this problem is the use of contributions plots, which show the contribution of each process variable to the calculated statistic. Although some strategies are proposed to assign variable-contribution values to the D2 statistic taking into account the multivariate nature of process data, it is still an open research subject. In this work, a new strategy to decompose the D2 statistic as a unique sum of each variable contribution is presented, in the space of the original variables. Also an explanation is provided regarding the meaning of negative contributions to the statistic. The results of the proposed strategy are compared with those obtained in the latent variable space by Westerhuis et al. (2000). The new decomposition avoids the lost of information originated by theprojection into a less dimensional space, that may lead to detection faults and false alarms.