Forecasting industrial aging processes with machine learning methods

Accurately predicting industrial aging processes makes it possible to schedule maintenance events further in advance, ensuring a cost-efficient and reliable operation of the plant. So far, these degradation were usually described by mechanistic or simple empirical prediction models. In this paper, we evaluate wider range data-driven models, comparing some traditional stateless models (linear kernel ridge regression, feed-forward neural networks) more complex recurrent networks (echo state LSTMs). We first examine how much historical data is needed train each on synthetic dataset with known dynamics. Next, are tested real-world from large scale chemical Our results show that produce near perfect predictions when trained larger datasets, maintain good performance even smaller datasets domain shifts, while simpler only performed comparably datasets.