WORKSHOP ON DEPENDENCE ANALYSIS
26 and 27 of SEPTEMBER, 2025
Institute of Mathematics, Statistics and Scientific
Computation of the University of Campinas, Brazil
SATELLITE WORKSHOP OF
8th BRAZILIAN CONFERENCE ON STATISTICAL MODELING IN INSURANCE AND FINANCE
Title: Sensor Monitoring Towards Industry 4.0: Bridging Concepts Apropos the cn Multivariate Control Chart
The biggest challenge in monitoring multiple sensors is related to the high dimensionality that a process can easily obtain. Additionally, Statistical Process Control (SPC) techniques have difficulty in simultaneously detecting change of quality characteristics when extrapolated from the normality (or Gaussian) assumption. That is, the most known SPC multivariate control chart is based on Hotelling's T2 statistic. Moreover, it is assumed that the process data follow a multivariate normal distribution (which in practice rarely occurs) and marginals are supposed to be uncorrelated. This study focused on the SPC multivariate case (dimension d>=2), regardless of the data dependence or marginal asymmetry. Thus, we proposed a new control chart for embedding change detection in the multivariate process, called the cn-Chart, which is analogous to the tolerance region based on the density level set from a copula function. Furthermore, this methodology quantifies the correlation (not necessarily linear) between the response variables. Two study cases were adopted: one investigated the water quality control task, and the other the brass-steel thermostats monitoring relationship. Our results show that both association and anomaly detection are captured using the proposed methodology, which also allows the inclusion of the variable selection as conditional statistical modeling. (Joint work with Diego Carvalho do Nascimento (UDA-Chile), Francisco Louzada (ICMC-USP and Ana Claudia da Silva Batista (Dest-UFBA).