Abstract

Abstract A milestone of Industry 4.0 is the improvement of the design procedures requiring models of complex processes. Models can be used to simulate the process, being accurate even if complex, and to predict process behaviour for control action, requiring simplicity and stability. In the last years, machine learning approaches came up alongside of the standard identification techniques for prediction purposes. In this work we propose two models of an industrial autoclave to describe the evolution of temperature and pressure. The first model (PhM) involves a physical structure with data-driven adaptation of the parameters, the second one is a Long Short-Term Memory network (LSTM), trained ensuring Input-to-State stability. Both models obtained good performance: FIT of 94.26% (91.55%) for the temperature (pressure) with PhM; 84.59% (78.31 %) for the temperature (pressure) with the LSTM. Future developments involve the synthesis of an MPC based on the LSTM to be tested in simulation via PhM.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call