Statistical Data Mining For Nominal Value Variance Anomaly With Visualization

A Nominal Value Variance (NVV) anomaly is an analog measurement whereby the measured value is within the “passing” range for a given test but has departed from the nominal value significantly such that a problem may exist. The occurrence of this anomaly could indicate a test implementation problem if the anomaly is occurring during Test Program Set (TPS) development. Manual discovery of a NVV anomaly is very time consuming when a large number of measurements are being made. Development and integration of new test equipment with a suite of new avionics hardware within an aggressive schedule and budget compounds the problem and typically means very little or no attention is given to address this anomaly. Often it is not until the TPS has been delivered to the customer and exposed to a large avionics equipment sample that a NVV anomaly is uncovered. This paper introduces a methodology to analyze measurements collected during test program execution to detect a NVV anomaly. The methodology is formulated by applying knowledge discovery techniques, also referred to as data mining, during the development effort to “mine” measurement records and visualize the results coherently as to ascertain whether an anomaly exists before the product is delivered to the customer. However, this methodology is predicated upon the fact temporal measurement records have been recorded and collected during the development effort. Currently, full coverage measurement data collection methods do not exist for military avionics test equipment development environments. To address this issue this paper identifies how the test measurement system “idle” time can be utilized to automate measurement recording and storage in standard data formats.