An artificial neural network approach to recognise kinetic models from experimental data

Abstract

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Time series analysis and long short-term memory (LSTM) network prediction of BPV current density

“…There are, of course, some trade offs when using artificial neural network models. First, it is difficult to assign physical significance to neural network parameters 22 . Second, since neural networks are typically trained on a set of experimental data, it is challenging to extrapolate beyond the conditions used in the experiments 22 .…”

“…First, it is difficult to assign physical significance to neural network parameters 22 . Second, since neural networks are typically trained on a set of experimental data, it is challenging to extrapolate beyond the conditions used in the experiments 22 . Therefore, robust neural networks may require large amounts of experimental data to be collected.…”

Safe model-based design of experiments using Gaussian processes

“…The parameter of the kinetic model can be seen as calibration parameters and estimated through their posterior distribution, which provides a non-asymptotic approach for uncertainty quantification. Additionally, the structural mismatch can be potentially be diagnosed [48] and different structure could be potential be found using artificial neural networks [49]. and second order Taylor for the variance Σ i (•, •):…”

Safe model-based design of experiments using Gaussian processes

“…The parameter of the kinetic model can be seen as calibration parameters and estimated through their posterior distribution, which provides a non-asymptotic way for the uncertainty quantification. Additionally, the structural mismatch can be potentially be diagnosed [40] and different structure could be potential be found via artificial neural networks [41].…”