Um Assistente Virtual para Suporte a Assistência Técnica de Bombas Centrífugas em Processos Industriais

Abstract

Research in automatic fault diagnosis using Artificial Intelligence (AI) techniques has been ongoing for decades, and more recently, methods like Machine Learning have improved the predictive capabilities of industrial systems. However, these advancements do not eliminate the need for technicians to complement maintenance work, whether in replacing components, performing preventive maintenance, or carrying out other maintenance actions. An effective maintenance process requires a well-trained team, established procedures, and supporting systems. Chatbots and virtual assistants have demonstrated success in human interactions across various fields. The present work explores this resource and presents the development of a Chatbot-type virtual assistant to support the technical team maintaining centrifugal pumps in an industrial context. The application aims to assist with the specificities of these devices, as well as in field removal and installation operations. The development was carried out with the Chatterbot tool in the Python language. Specialized technical knowledge was modeled and embedded in the application. Machine Learning resources in Chatterbot were applied to train and refine user responses. Preliminary tests with a reduced database indicated that the E-MANUEL chatbot achieved a success rate of 89.2%, and in only 7.7% of interactions it was unable to provide a response to the user within the expected context.