Method and apparatus for predicting object properties and events using similarity-based information retrieval and modeling

Computer apparatus for the dynamic indexing of spectral data, the spectral data being of a spectral database tree having a root node, the computer apparatus comprising search manager apparatus, the search manager apparatus comprising a processor, memory associated with the processor and an output device, the computer apparatus comprising

• a memory for storing the spectral data of the spectral database, the spectral data being represented by an attribute vector haying a plurality of M dimensions, where M is at least three, the spectral data comprising one of electromagnetic, acoustic, optic, isotopic, mass spectra and thermal data;
• the processor for defining a root node of the spectral database tree and for performing a test for the root node responsive to a query received at the processor search manager apparatus, the query related to an object, the test defining an indexing surface in a three dimensional sub-space for separating the spectral data into at least two clusters of the spectral data;
• the processor for determining reference evidence data from the spectral database responsive to the query via the output device, the query for one of a location, a manufacturer and an event related to the query object, the query object being a target object;
• and
• displaying the target object at a location central to a display, the target object displayed as a symbol and linked by one of a line and a curved line to another object, one of the lines forming part of a pathway having a similarity value, a speculation level and a model-based rank, the rank being an ordinal value balancing the similarity value with the speculation level.

A computer-implemented method of determining one of a worn component and a failed component in one of a machine and a process via inputting a query to a processor search manager apparatus, the processor search apparatus being coupled to a plurality of databases and comprising a processor and an associated memory, the plurality of databases comprising a spectral database, the processor connectable to a data acquisition device and to an output device, the computer-implemented method comprising:

• storing data objects is said spectral database, each data object comprising spectral data received in memory via said connectable data acquisition device, the data acquisition device being coupled to the one of the machine and the process;
• storing at least one attribute associated with the spectral data of the one of the machine and the process in a second database of the plurality of databases, the second database being different from the scctral database;
• processing a query comprising information about a target object, the query information containing spectral data;
• using said search manager apparatus to search the spectral database;
• the search manager apparatus applying multivariate statistical analysis and clustering to determine a tree-structured index for indexing and searching the spectral database to retrieve data objects most similar to the target object;
• retrieving stored information from said spectral database about the retrieved data objects having similar spectra according to a similarity measure;
• retrieving attribute information associated with said retrieved stored information from said second database;
• and
• using a model, the retrieved stored information of the data objects having similar spectra and the retrieved attribute information from the second database to predict at least one property of the target object in response to the processed query, the predicted at least one property determining one of the worn component and the failed component in the one of the machine and the process for output via the output device.

A computer-implemented method of determining one of a worn component and a failed component in one of a machine and a process via inputting a query to a plurality of databases coupled to processor search manager apparatus, the processor search manager apparatus comprising a processor and associated memory, the processor for outputting data via an output device, the computer-implemented method comprising:

• selecting a central object associated with one of the machine and the process, the central object having data stored in at least one of a plurality of databases of the processor associated memory;
• determining via the processor and outputting data for display via the output device an evidence tree for the central object, said displayed evidence tree representing a decomposition of said central object into at least one component part and at least one associated evidence specimen;
• determining a set of objects similar to at least the one associated evidence specimen using a similarity measure and a threshold, the determining comprising applying multivariate statistical analysis and clustering to determine a tree-structured index for indexing and searching one of the plurality of databases to retrieve the object most similar to the central object;
• outputting at least one of the set of objects for display via the output device and a connection of said at least one object to the associated evidence specimen;
• applying at least one model to predict a property of the component part of the central object using the set of objects;
• and
• using the predicted property to determine and output an identity of the one of the worn component and the failed component in the one of the machine and the process.

A computer-implemented method of determining one of a worn component and a failed component in one of a machine and a process via inputting a query to a plurality of databases for storing objects, the plurality of objects coupled to a processor search manager apparatus, the processor search manager apparatus comprising a processor and memory, the processor search manager apparatus for coupling to an output device, the computer-implemented method comprising:

• searching via said processor search manager apparatus each of at least two databases for stored objects having attributes similar to at least one selected attribute of a target object, said selected attribute of the target object being different for each of the at least two databases, the searching by applying multivariate statistical analysis and clustering to determine a tree-structured index for indexing and searching the at least two databases to retrieve the objects most similar to the target object;
• the two databases comprising a frequency spectral database and a thermal signature database for each object;
• retrieving attribute values associated with said stored objects in each database of said at least two databases using a similarity measure;
• and
• determining a predicted object property of the target object for output via said output device using said retrieved attribute values and a model comprising a function of at least one of the retrieved attribute values, the predicted object property of the target object being associated by at least a linkage with a similar object of one of the frequency spectral database and the thermal signature database.