A case study in on-line intelligent sensing
A.W. Moran, Paul G. O’Reilly, G.W. Irwin · 2000
A new method is described for online detection of parameter changes in a sensor. This is based on work by Yung and Clarke (1989) which employs a local ARIMA model of the sensor output to generate an innovation sequence. A statistical test, which quantifies the change to the variance of an innovation sequence, is developed and used to provide a decision process based on a likelihood ratio of probabilities. Real-time experimental results for detecting a change in a thermocouple time-constant are presented.