Fast online identification of low-order time-delayed industrial processes

Although online identification of low-order industrial processes is straightforward when the time delay is known and fixed, the problem becomes significantly more challenging when the delay is unknown and/or varying. Presented is a fast, exact recursive identification algorithm for this situation with a memory and time complexity that is linearly bound in the unknown delay. Experiments indicate the method significantly improves upon the memory and computation requirements typical of previous online methods, and possesses improved convergence properties.