Automated Simulation Error based Reduction (ASER) of large chemical reaction mechanisms. (2nd November 2019)

• Record Type: Journal Article
• Authors: Veerappan, Devi Raghavee; Ramanathan, Karthik; Kaisare, Niket S.
• Is Part Of: Computers & chemical engineering. Volume 130(2019)
• Publication Date: 2019-11-02
• Subjects: Mechanism reduction -- Principal Component Analysis -- Directed Relation Graph -- Simulation error -- Semiconductor manufacturing -- Chemical vapor deposition; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2019.106560
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: An Automated Simulation Error Based Reduction (ASER) approach is proposed. ASER is implemented with Directed Relation Graph and Principal Component Analysis. ASER-DRG-PCA is used to reduce gas-phase H2 –O2, and GaAs deposition mechanisms. Reduced mechanisms with simulation error below the specified limit are generated. Establishes relationship between simulation error and threshold parameter for reduction. Abstract: We propose an Automated Simulation Error based Reduction (ASER) algorithm for determining reduced kinetic models. It directly uses simulation error between detailed and reduced mechanisms for determining the final reduced model. ASER is implemented using Directed Relation Graph (DRG) for species reduction and Principal Component Analysis with concentration sensitivity (PCA-S) for reaction reduction. The thresholds for DRG importance index, and the cutoffs for eigenvalues and eigenvectors in PCA determine the final reduced mechanisms. However, the effect of these thresholds on the reduced mechanism is non-intuitive, as they are not explicitly correlated with the simulation error. ASER indirectly establishes this correlation by starting with large values of thresholds and updating them iteratively until reduced mechanism that meets the user-specified simulation error is obtained. A combination of ASER–DRG followed by ASER–PCA is shown to yield an optimally reduced mechanism. The proposed algorithms are tested using gas-phase H2 -O2 and gas/surface GaAs deposition mechanisms.