Knowledge Extraction from Catalysis of the Past: A Case of Selective CO Oxidation over Noble Metal Catalysts between 2000 and 2012

Abstract

The objective of this work is to demonstrate that some valuable knowledge can be extracted from past publications by using various data mining tools so that the continuously growing experience accumulated in the literature over the years can be used in a more effective manner. Selective CO oxidation over noble metal catalysts is chosen as a case to test the validity of this approach because a considerable number of papers were published on this subject in the last decade. Thus, 249 papers published in the last 12 years have been inspected, 80 of which were used to form a database containing 5610 data points. First, the database was analyzed by using decision tree classification to determine the conditions that lead to high CO conversion. Then, the relative importance of various catalyst preparation and operational variables for CO conversion were determined by using artificial neural networks. Finally, the database was separated into smaller clusters by using a genetic algorithm‐based clustering technique, and the data in each cluster was modeled by artificial neural networks to predict the effects of individual catalyst preparation and operational conditions on the catalytic activity. All these analyses were effective in the extraction of knowledge from the literature and the deduction of some useful trends, rules, and correlations, which are otherwise not easily comprehensible.

Statistical Analysis and Discovery of Heterogeneous Catalysts Based on Machine Learning from Diverse Published Data

“…For quantitative evaluation, three published datasets of heterogeneous catalysts and their performances were analyzed (Table ) for OCM, WGS, and CO oxidation reactions . The original OCM dataset consisted of 1866 catalysts with information on their compositions, support types, promoter types, experimental conditions (preparation methods, operating temperature, total pressure, and contact time), and reported performance (yield and selectivity of C 2 hydrocarbons).…”

Statistical Analysis and Discovery of Heterogeneous Catalysts Based on Machine Learning from Diverse Published Data

“…For quantitative evaluation, three published datasets of heterogeneous catalysts and their performances were analyzed (Table 1) for OCM, [47] WGS, [51] and CO oxidation reactions. [58] The original OCM dataset consisted of 1866 catalysts with information on their compositions, support types, promoter types, experimental conditions (preparation methods, operating temperature, total pressure, and contact time), and reported performance (yield and selectivity of C2 hydrocarbons). To ensure the quality of the statistical data, the following preprocessing was applied.…”

“…# Catalysts (Original) # Features Target OCM [47] 1833 (1866) 105 C2 yield (%) WGS [51] 4185 (4360) 61 CO conversion (%) CO oxidation [58] 5567 (5610)…”

Reaction condition optimization for non-oxidative conversion of methane using artificial intelligence

“…12 ML can match input variables (or features) to target properties and thus can be generally employed to available datasets for numerous catalytic reactions. 4,[13][14][15] For example, successful applications of ML methods toward the optimization of heterogeneous catalysts are shown for a database of oxidative coupling of methane (OCM) which contains ~1800 catalysts and reaction conditions. 4 With the advancements in data science, chemical reaction data can be investigated with several approaches such as statistical analysis, 4 machine learning.…”