Learning based Probabilistic Model for Migration of Industrial Control Systems

• Published in: Proceedings (IEEE International Conference on Emerging Technologies and Factory Automation) pp. 87 - 94
• Main Authors: Mohan, Devina, Jetley, Raoul, Sudarsan, SD
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2019
• Subjects: C++ languages; Control systems; Learning System; Machine Learning; Migration; Naive Bayes; Ontologies; Probabilistic logic; Probabilistic Model; Software; Suggestion Management System; Task analysis
• ISSN: 1946-0759
• DOI: 10.1109/ETFA.2019.8869172

The updating and upgrading of control systems is a cumbersome, expensive and time consuming task. From a software perspective, control system migration is a collective task of migrating the control logic, Human Machine Interface (HMI) and auxiliary software applications. Migrating control logic is the most challenging task owing to constraints on hard real-time behavior and execution order. Control logic typically contains engineering artifacts that specify the functionality of industrial devices taking into account various parameters. Therefore, to migrate from one Distributed Control System (DCS) system to another or to upgrade the existing DCS, one needs to map the source control entity and their parameters to the appropriate control entities in the target DCS.In this paper, we propose a machine learning based suggestion management system that identifies control entities and parameters for a source DCS and suggests the use of similar control entities and corresponding parameters for the target DCS. This in effect saves effort required in mapping of control parameters and reduces the dependence on subject matter experts. Our system uses a probabilistic approach to find these similarity mappings based on meta-data stored in an Ontology. We further describe a case study implemented for mapping heritage and legacy systems to a modern control system to verify and validate our approach.