Turboshaft Engine Weight Estimation Model Development Based on Statistical Data
by HAVADER Editör Ekibi
Imagine trying to estimate the price per square meter of a house that hasn't been built yet, with only historical data from similar projects to go on. Turboshaft engine designers face a similar problem: they have to estimate an engine's weight before it's built, based only on design parameters.
The goal of this study was to develop a model that can estimate targets like weight, which can't be determined analytically during a jet engine's preliminary design phase. The researchers built a database from existing engine data in the literature and developed a dedicated turboshaft engine weight model using regression in MATLAB, splitting engines into light-weight and heavy-weight categories to improve accuracy.
The results came with statistical parameters showing how well the model reflects reality: for light-weight turboshaft engines, a model built on 48 engine cases yielded a maximum error of about 40%; for heavy-weight engines, a model built on 16 cases yielded about 30%. Those numbers might look high at first glance, but they represent an honestly reported finding: the model's ability to represent reality is limited, and that's stated as a natural consequence of the relatively small sample size and the varying technology levels of the engines in the database.
What this study contributes isn't so much a "successful" model as an honest, transparent one. An everyday analogy: it's similar to a real estate agent saying "homes in this area sell for roughly this price, but the sample is small, so expect a 30-40% deviation" — rather than giving a false sense of precision, it openly states the estimate's confidence bounds.
In the end, this research gives aviation engine designers a concrete reference for how much uncertainty to factor in when estimating weight at the preliminary design stage — helping set more realistic budgets and performance targets.
← Blog
The goal of this study was to develop a model that can estimate targets like weight, which can't be determined analytically during a jet engine's preliminary design phase. The researchers built a database from existing engine data in the literature and developed a dedicated turboshaft engine weight model using regression in MATLAB, splitting engines into light-weight and heavy-weight categories to improve accuracy.
The results came with statistical parameters showing how well the model reflects reality: for light-weight turboshaft engines, a model built on 48 engine cases yielded a maximum error of about 40%; for heavy-weight engines, a model built on 16 cases yielded about 30%. Those numbers might look high at first glance, but they represent an honestly reported finding: the model's ability to represent reality is limited, and that's stated as a natural consequence of the relatively small sample size and the varying technology levels of the engines in the database.
What this study contributes isn't so much a "successful" model as an honest, transparent one. An everyday analogy: it's similar to a real estate agent saying "homes in this area sell for roughly this price, but the sample is small, so expect a 30-40% deviation" — rather than giving a false sense of precision, it openly states the estimate's confidence bounds.
In the end, this research gives aviation engine designers a concrete reference for how much uncertainty to factor in when estimating weight at the preliminary design stage — helping set more realistic budgets and performance targets.