AI-Powered Automation Of BSS Operations In Manufacturing Ecosystems: A Cloud-Native Approach

Abstract

Manufacturers are continually facing significant challenges in their pursuit of growth and innovation, particularly amid ever-increasing cost pressure, intense competition, and accelerated digitization across the industry. The widespread adoption of data acquisition tools facilitates massive data collection capabilities; however, despite this potential, innovations tend to be slow and costly. As a result, they often fail to effectively reach the shop floor because of inefficient sharing mechanisms and the problem of scattered data being stored across different platforms. The design of systems within manufacturing environments relies heavily on the accumulated experience of engineers and managers, coupled with time-consuming simulation models. This approach [1]ultimately restricts their ability to manage complexity efficiently. In addition to these challenges, the necessity for reliable predictive maintenance (PdM) solutions is becoming ever more critical in the industry. At the same time, without standardization of data formats and practices, providing accurate and consistent data for the higher echelons of the supply chain becomes increasingly complex and problematic, further complicating these issues.

The growing interest in AI methods for data-intensive process-driven data enables untapped knowledge from already available massive data to be converted into information or action. A data-centric solution approach tackles the data sharing issue, delivering data-driven frameworks and developing reference solutions for boosting the uptake of AI modeling. This stands for the design of a reference data mart for a batch production use case. Stochastic warping is a promising method for operating under high uncertainty data processes as factory ecosystems. However, its performance strongly depends on an exact sampling of the simulation environment response. Uncertainty quantification and its feasible quasi-random full factorial sampling approaches for weight dimensions and continuous inputs are suggested.