Revisiting Legacy Weight Relationships Using Machine Learning Techniques

Abstract

This paper investigates the application of K‐means clustering algorithms to traditional aircraft conceptual‐level weight estimation techniques. K‐Means clustering was utilized as a means for statistical discovery within the dataset to explore potential paths for improved weight estimation using conceptual‐level information. As a proof‐of‐concept demonstration, scope was limited to fuselage basic weight estimation. A variety of weight sources were parsed and curated to produce a large, diverse dataset consisting of 82 separate aircraft. A corresponding new universal fuselage basic weight regression was generated as a baseline for comparison. K‐Means clustering was then employed to sort aircraft into groupings based on configuration and topology with an associated weight equation created for each grouping. Configuration-based groupings utilized information such as a high-level abstraction of the structural layout as well as whether the aircraft is a fixe‐wing or rotary‐wing vehicle. Topology‐based groupings utilized information such as landing gear location and possession of a cargo ramp or wing. The configuration‐based groupings showed modest improvement compared to the baseline regression which were in turn outperformed by the topology-based regressions across a range of data groupings. Under all conditions, a subset of the data associated with fixed‐wing aircraft was shown to be an outlier in regard to error as a result of a large range of weight and speed scales, in addition to possible secondary pressurization impacts. Special treatment of the winged dataset led to further reduction in error based on unique design features. The weight estimation processes explored within this paper present a methodology that leverages machine learning algorithms that can improve and inform existing best practices.