DIPy-AI: Brain-Cognition-Inspired DIKW Pyramid-Based Agile AI Architecture for Industrial Sensor Data Assimilation

The paper proposes DIPy-AI, an agile AI architecture based on the data-knowledge-information-wisdom (DIKW) pyramid, for processing sensor data in production environments. DIKW is one of the accepted models abstracting the assimilation of sensory data by the human brain. DIPy-AI aims to address challenges related to data assimilation, quality detection, and modular information extraction. The proposed architecture consists of three layers, viz a sensor-dependent data pre-processing layer, a sensor-agnostic ML layer for converting data into information, and an application-specific layer for knowledge extraction. There are two major merits of the proposed architecture. By having a layered architecture, it can easily be repurposed for different industries. Secondly, this agility in the architecture also facilitates the changing of sensors as well as overall goals of the architecture. The work aligns well with sustainable industrial digitization goals (shared by many countries) and offers a flexible solution applicable to multiple industries, promoting sustainability, data-sharing and architecture sharing.