The development of a model for electrodialysis enables us to predict the influence of process setting on the deterioration of the process. Physics-based models are often used for this purpose but in many cases, these models are very complex, have a very long development period and notorious for their computational expenses. Machine learning is a time and computationally efficient way of obtaining a process model at the cost of more data and less extrapolative power. In this presentation, we explore the use of artificial intelligence (AI)(vervang door machine learning voor een meer wetenschappelijkere tekst;) in the form of black-box differential equations. In this approach, we define a neural network as a differential equation to facilitate time-dependent predictions while providing a great trade-off between data need and extrapolative capabilities. In this presentation, we apply this technique to predict the deterioration of an ED process heave fouling.
#process #electrodialysis #modeling #neuralnetworks
The slides of this presentation can be found here: