Abstract
A previously conducted study has developed a comprehensive framework that combines Machine Learning techniques and environmental assessment to facilitate decision-making in Process System Engineering (Prioux et al., 2023). This framework consists of five distinct steps: goal and scope definition, data architecture establishment, sustainability assessment, data visualization and result analysis, and the final decision-making process. The data architecture itself is directly influenced by the construction of big data architecture and comprises five sub-steps: (i) data collection and extraction from Knowledge Engineering, (ii) data enrichment and storage facilitated by Process Engineering and Machine Learning (ML), (iii) data cleaning, (iv) analysis of (raw) data, and (v) visualization of (raw) data. In the third step, environmental impacts are calculated, while the fourth step involves the utilization of ML tools for analyzing and visualizing the obtained results. To evaluate the efficacy of this approach, a comparison of various biomass pretreatment processes for glucose production was conducted. The primary objective of this study is to examine the steps of data enrichment and data cleaning in detail. To obtain relevant scientific articles, a targeted search using specific keywords was conducted on platforms such as Science Direct and Web of Science. However, it was observed that the data quality among these articles is inconsistent, with the presence of potential inconsistencies. Leveraging Machine Learning (ML) techniques enables us to enhance the quality of the data. This research paper seeks to emphasize the significance of ML, particularly neural network tools, in the context of environmental assessments within the realm of Big Data. The key focal points of our approach encompass enriching data from literature sources to simulate processes and identifying outliers within the extracted data.
