Dr. N. Jordan Jameson (CALCE Alumnus) and Dr. Michael H. Azarian are co-inventors on a patent that was recently granted by the United States Patent and Trademark Office, “Systems and Methods for Determination of Health Indicators Using Rank Correlation Analysis.” (https://patents.google.com/patent/US11086750B2/en)

To perform system prognostics and health management (PHM), one or more sensitive indicators are needed to characterize the system's health state. When presented with multiple potential system health indicators, it is necessary to decide which sensor data and/or extracted features are sensitive to system health.

In the past, multiple sensor data and/or features were combined together using dimensionality reduction techniques like principal component analysis (PCA) or Mahalanobis distance, which can be computationally expensive and obscure the physical interpretability of the analysis.

The patented method uses rank correlation analysis to quantify the extent to which a given health indicator tracks with system aging or degradation time. This type of ranking allows engineers to reduce the number of sensors and amount of data needed for health monitoring, enabling them to focus just on those features that are best suited to serve as health indicators

There are many potential applications of this method for sustainment and failure prevention of products ranging from components or assemblies to complex systems. For example, it has been applied at CALCE to the non-invasive health monitoring of electrical coils, whose degradation has traditionally been difficult to track and predict. Impedance spectroscopy has been found to be a useful, non-invasive method for collecting health-related information on coils used in applications such as solenoid valves, motors, relays, and transformers. The problem arises with the large number and range of frequencies that are contained in such spectra, making it challenging to extract insulation health from the raw impedance data. It would be impractical to record impedance values at so many different frequencies and to process the data in order to extract health information in a fielded application.

Fortunately, some frequencies better reflect the aging condition of the coil than others. One measure of this relationship is the monotonic trends of impedance measurements at certain frequencies over the life of the product. By performing some trial tests and applying rank correlation analysis to the data, a handful of the most informative frequencies can be identified, which can then be used to develop a cost-effective and accurate PHM solution.

For more information on the patent and potential applications such as electrical coil health monitoring, please contact Dr. Michael H. Azarian (mazarian@umd.edu).