Data-Driven Prediction of Polymer Intrinsic Viscosity with Incomplete Time Series Data.

Prediction of polymer intrinsic viscosity is very important for stabilizing the quality of polyester fibers. Data-driven prediction has been applied to solve engineering problems. However, incomplete data are all too frequent to a phenomenon in the workshop due to signal interference, human error, and other reasons. This paper proposes a data-driven approach for polymer intrinsic viscosity prediction with incomplete series data. First, a time series data generative adversarial network (TSDGAN) is presented to fill in the missing data, in which Attention LSTM serves as the generator and CNN as the discriminator. Next, An Informer model is introduced for polymer intrinsic viscosity prediction with the generated time series data. According to the experimental results under different missing rates, the proposed approach achieves better results than traditional and some state-of-the-art time-series prediction methods.