Fei Comapny

Michael Buhot

Michael James Owen

Van Hung Phan

The present invention discloses a combination of two existing approaches for mineral analysis and makes use of the Similarity Metric Invention, that allows mineral definitions to be described in theoretical compositional terms, meaning that users are not required to find examples of each mineral, or adjust rules. This system allows untrained operators to use it, as opposed to previous systems, which required extensive training and expertise.

Automated mineral classification

Mineral definitions each include a list of elements, each of the elements having a corresponding standard spectrum. To determine the composition of an unknown mineral sample, the acquired spectrum of the sample is sequentially decomposed into the standard spectra of the elements from the element list of each of the mineral definitions, and a similarity metric computed for each mineral definition. The unknown mineral is identified as the mineral having the best similarity metric.

Mineral identification using sequential decomposition into elements from mineral definitions

An improved mineral analysis system includes mineral definitions that include not only characteristics of the minerals, but also variability in those characteristics. The variabilities allow the calculation of ranges of expected values for different quality of measurements, for example, for different numbers of x-ray counts. Match probabilities can therefore be calculated to more accurately determine the composition of a mineral sample.

Mineral identification using mineral definitions including variability

The present invention discloses a method for determining the mineral content represented by the entire SEM-EDS dataset, including initially unknown data points. SEM-EDS data points are taken and compared to a set of known data points. Any data point that is not sufficiently similar to the known data point is classified as unknown and clustered with like unknown data points. After all data points are analyzed, any clusters of unknown data points with a sufficient number of data points are further analyzed to determine their characteristics. All clusters of unknown data points with an insufficient number of data points to allow further analysis are considered outliers and discarded.

Growing community of inventors

Brisbane, Australia

Michael Buhot