[video]
Presenters: Byron Biggs and Sergiy Peredriy
KT chart monitoring enables rapid anomaly detection and process control for high-frequency multivariate data. The method is based on support vector data description (SVDD) applied to a moving window of observations. KT charts are non-parametric sub-group control charts, which can simultaneously monitor central tendency and spread. The method accommodates non-normal data and supports monitoring multimodal processes. Potential industrial IoT applications include predictive maintenance and predictive quality applied to high-value industrial assets and processes. KT chart monitoring consists of two steps: 1. Training – The KTTRAIN procedure in SAS Viya uses the data from normal operations to define the state of statistical control. KT chart training calculates the control limits for permitted deviations in the central tendency and the spread of the process. The KT chart for monitoring the process center is called the A chart, and the KT chart for monitoring the process variation is called the R2 chart. 2. Monitoring – The KTMONITOR procedure in SAS Viya and Streaming KT Chart Monitoring in SAS ESP use the data from the ongoing process to monitor for stability in both the central tendency and the spread by using the control limits that are calculated during the training step. This super demo showcases the capabilities of KT chart monitoring applied to industrial process data. It includes the training process and the detection of anomalies with streaming data.
