A time-varied predictive model for EDM process

The classification techniques of discharging pulses in EDM have been proved critical in improving productivity, precision and lowering the cost of products, etc. In this paper, an easily implemented method is developed to describe the variations of EDM process, represented by gap states. On the basis of a time series of gap states from a machining process, the paper first studied a general descriptive model for EDM process, and then equivalently simplified the model for application; after spectral analysis, preprocessing of data, parameters selection and model validation proposed a well-defined model. Finally, by using this model structure and size, an online time-varied predictive model was developed. Experimental verifications showed that this predictive model can quickly and accurately provide one step ahead predictions with mean error less than 2%. This model makes clear that variations of EDM process represented by gap states can be predicted online with a high precision.