Invention Reference Number
This invention is directed to a machine leaning methodology to quantify the association of a set of input variables to a set of output variables, specifically for the one-to-many scenarios in which the output exhibits a range of variations under the same replicated input conditions. The scenarios arise in many industrial processes and in various experiments in which repeated measurements of an output quantity vary under the same, often complicated, experimental conditions, and the degree of such variations of output quantity is influenced by the changes in experimental conditions.
The invention utilizes a multichannel (or segmented, or other terms better descripting our approach) neural network system to quantify and assess such one-to-many variations. It further provides a means to associate of output variables to the input variables, and quantify the consistency of outcome under different input conditions. It has broad usage, including to evaluate the product performance and consistency in manufacturing, and determine optimal process parameters tailored to various materials, meanwhile capturing the variability inherent in the manufacturing process.
Contact
To learn more about this technology, email partnerships@ornl.gov or call 865-574-1051.
