Industrial automation is evolving by leaps and bounds due to digitalization efforts, and this is helping to create an adaptable Industrial Internet of Things. This transformation of manufacturing facilities into large-scale systems requires addressing the complicated issue of making heterogeneous systems and their data models and standards compatible, which is a major issue in constructing service-oriented automation systems. This thesis deals with the problem of establishing automated service interoperability when exchanging heterogeneous message data. A machine-learning architecture is developed to optimize message transcoders and system of systems utility to establish interoperability. By optimizing transcoders dependent on both service and metadata, the objective is to ground the learned latent representations into the physical environment for better generalization.
Two experiments using physical simulations were completed to examine and evaluate the architecture for generating heterogeneous JSON messages from multiple heating and air conditioning services. The first experiment applies unsupervised learning with back-translation for transcoding created features from message services, attaining an accuracy of 49% for message translation. The second experiment focuses on supervised learning with a modular neural network (JSON2Vec) for automated encoding of the heterogeneous JSON messages, providing the capacity for correct interpretation of messages in terms of the anticipated system of systems behaviour. These results demonstrate that machine learning is a viable path for interoperability automation research, which will benefit from both symbolic metadata and message data for generalization and adaptation. Datasets open to the public are needed to develop this kind of research further, as well as to move solutions to automation systems.
When talking about industrial automation, the person highlighted in this article is Professor Fredrik Johansson with the Department of Electrical Engineering at Chalmers University of Technology. His expertise in topics such as machine learning and its application to service data interoperability makes him an invaluable asset to the industry.
The company mentioned in the article is Chalmers University of Technology. It is a renowned university located in Gothenburg, Sweden. It was established in 1829 and is one of the oldest, most prestigious universities in the country. With its commitment to providing advanced education and research today, it has become a leader in the fields of engineering, technology, and design. It also offers international programmes for aspiring masterminds throughout Europe.
