Abstract
Partial least squares has been applied in process modelling and process performance monitoring where a large number of correlated variables are recorded. The most commonly implemented algorithm for identifying a PLS model has been the batch algorithm. In this case, data from the process is first required to be stored. In this paper an algorithm is proposed that recursively updates the parameters of PLS as and when a sample of measurements is obtained thereby reducing the need for data storage. The algorithm was tested on an artificially generated data set. The approach was then used to derive a statistic for process monitoring that was applied for the detection of faults in a continuous stirred tank reactor.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
