Building an EDM process model by an instrumental variable approach based on two interactive Kalman filters

An electrical discharge machining (EDM) process model is critical in control system for stabilizing EDM or micro EDM processes and improving productivity. This paper presents a new feasible method to build an online process model. A two order AR model was firstly proposed for the process, but the deduced formulation of the model parameter estimates is biased. An instrumental variable approach is then proposed to correct the biased estimates. This approach was completed by two interactive Kalman filters. One filter provides instrumental variables, gap state estimates, through the latest estimated parameters; while the other uses the gap state estimates to recursively estimate a two order auto-regressive (AR) model parameters and then provides a one-step gap state prediction. The two Kalman filters by way of interactively supporting each other forms a new feasible way of online modeling an EDM process.