Virtual Voice Assistant for Digital Twin of Production Facility

The paper considers the use of virtual voice assistants for interaction with industrial facilities and their digital twins. The authors describe the architecture and a set of technologies suitable for the development of voice assistants for production settings. The main focus of this research is the problem of classifying the text of commands obtained by voice recognition in noisy conditions. The authors used fuzzy string matching, string matching with cosine similarity, and machine learning models to recognize commands. Machine learning model was the most reliable command text recognition method of all those reviewed in this paper. The article also presents an example of voice assistant module implementation as part of the software package designed for the development of digital twins of energy facilities.