An Algorithm for Diagnosis of Aircraft Engine Icing Conditions

At the Research Applications Laboratory at the National Center for Atmospheric Research, a prototype system called ALPHA (Algorithm for Prediction of HIWC Areas) has been developed to diagnose High Ice Water Content (HIWC) conditions. The purpose of ALPHA is to evaluate weather conditions related to power loss events of commercial aircraft due to accumulation of ice in jet engines. Deep convective clouds have been a common factor in over 100 cases that create safety issues and cause expensive damage to the aircraft. ALPHA uses numerical weather prediction model, satellite, and radar data to create an hourly diagnostic field to highlight cloud conditions that are conducive to HIWC. The diagnostic field is created using an expert system based on scientific research that uses fuzzy logic to combine data and approximate their influence.

ALPHA is a data-driven system. It is primarily composed of C++ applications that have been tested and used in other software systems all over the world. These applications are organized in a chain to convert and combine data types, remap data to a common grid, derive interest fields, and create graphical outputs of the HIWC hazard. The system runs continuously, polling data directories for new data to process. ALPHA also includes a system monitoring capability and an interactive JAVA-based display.

Two instances of the prototype system monitor the CONUS and northern Australia. An iterative software development process is used to ensure that the introduction of new functionality is integrated properly with the rest of the system. Since ALPHA was structured to readily change the interest field calculation and test improvements, new data sets (such as lightning or additional satellite-based products) may be incorporated into the algorithm. ALPHA will be tested in a field campaign in Darwin, Australia, during which an instrumented research aircraft will fly through clouds to collect information about their microphysical properties and verify the accuracy and significance of the interest values calculated by the algorithm.

This research is in response to requirements and funding by the Federal Aviation Administration (FAA). The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA. Additional support was provided by a NASA ROSES grant.