Abstract
Process monitoring has gained significant attention in recent years due to the need for certain industry sectors to enhance their processes' performance and safety. This development has enabled, more than ever, novel applications in real industrial systems. However, the desire to achieve quick results has led to decentralized and unstable applications that can hinder the long-term scalability and maintenance of these technologies. Moreover, the proliferation of commercial tools in response to the industry's demand for digital transformation has made selecting the right solution a daunting task. To address these challenges, Petrobras developed the SmartMonitor platform. SmartMonitor empowers users to create and configure, usually within minutes, online machine learning and first-principles monitoring tasks using a user-friendly, no-code visual programming framework. This approach ensures accessibility and democratizes the process of task creation and management. Additionally, the platform supports the inclusion of new methodologies and monitoring techniques, making it an integrated development hub aligned with best practices in machine learning models management. Currently, SmartMonitor has hundreds of tasks running in real-time, generating performance indices on critical equipment in Petrobras units. This paper provides a description of some of these tasks, along with an overview of the SmartMonitor platform's structure, its monitoring philosophy and challenges regarding process monitoring in real industrial systems.