Manufacturing Data Compiler for Visualization Based on Engineering-Driven Machine Learning

There has been substantial development in data analytics. However, the complex mathematical formulation of “big data analytics” is difficult to populate in general manufacturing plants. There is a need for an SPC-like tool to enable the acceptance of the advanced data analytics and its visualization. The MD Compiler, a tool to bridge the gap between data available and the information demanded by the users, is proposed. It will employ a hierarchical approach along with machine learning to complement the knowledge of the process so that the manufacturing data streams generate the informative process indexes that can be formulated quickly and presented to the operators for the illustration of the intended process attribute(s) and/or causal relations. The innovation of this SBIR proposal consists of the integration of the industrial process/product states into the data pipeline and applying unsupervised machine learning to automate the “mapping” processes in the analytics along with the concept of process snapshots. The MD Compiler, once developed, is expected to be a software package with functional modules that can take in the source data and automatically compile the data into informative process indexes for presentation with computing and data acquisition hardware support.