Uncertainty Assessment using Monte Carlo Simulation in Net Thrust Measurement at AETF

Bo-Hwa Lee, Kyungjae Lee, Inyoung Yang, Soo-Seok Yang, Daesung LeeInternational Journal of Aeronautical and Space Sciences 2007 · DOI

Abstract

In aerospace testing, accurately measuring net thrust is critical for engine performance validation. This study evaluates the Monte Carlo Simulation (MCS) method as an uncertainty assessment tool for net thrust measurements at the Altitude Engine Test Facility (AETF). By analyzing uncertainty sources and their probability distributions, the researchers compared the MCS approach with the conventional sensitivity coefficient method. The results demonstrate that the MCS method, which accounts for the inherent non-linearity of engine model equations, provides a more robust assessment of measurement reliability. Specifically, the MCS method identified a higher uncertainty (1.209%) compared to the traditional linear approximation (0.906%), suggesting that standard methods may underestimate measurement error in complex aerospace systems.

The Problem: Linear Approximations in a Non-Linear World

Net thrust in jet engines is not a direct measurement but a calculated value derived from a complex network of sensors measuring pressures, temperatures, and airflows. Each sensor introduces its own margin of error, or uncertainty. Traditionally, the industry has relied on the "sensitivity coefficient method," which uses a first-order Taylor expansion to combine these individual uncertainties.

While this linear approach is mathematically convenient, it often fails to capture the true physics of jet engines, which are governed by non-linear equations. By oversimplifying the mathematical model, the traditional method risks providing an inaccurate—and often overly optimistic—view of measurement reliability. The researchers set out to see if Monte Carlo Simulation—a method that uses repeated random sampling—provides a more realistic picture.

Experimental Setup and Methodology

To address these limitations, researchers at the Korea Aerospace Research Institute (KARI) compared the traditional method against Monte Carlo Simulation (MCS) using data from a single-spool turbojet engine. The test was conducted under sea-level conditions at an inlet Mach number of 0.7. The physical setup of the engine and its measurement components at the AETF is illustrated in Figure 1.

Schematic of the engine test setup at KARI's AETF, highlighting the engine, inlet duct, and thrust measurement frame.

The MCS approach bypasses complex partial derivatives by repeatedly "running" the thrust equation thousands of times, each time using random values drawn from the probability distributions of the input sensors. This "brute force" statistical sampling naturally incorporates the non-linearities that Taylor expansions ignore.

Ensuring Simulation Stability

A key challenge in MCS is determining the number of iterations required to produce a statistically stable result. The study analyzed the distribution of total pressure ($P_{T,g}$) across varying simulation counts. As shown in Figure 2, the distribution stabilizes and approaches a Gaussian curve as the number of simulations increases.

Histograms of total pressure ($P_{T,g}$) for 50, 500, and 5,000 simulations, demonstrating the convergence of the MCS method.

The researchers concluded that 5,000 simulations were necessary to ensure that statistical noise remained below 1%, providing a robust foundation for the final uncertainty calculation.

Key Results and Insights

The comparison revealed a significant gap between the two assessment methods:

  • Traditional Method Uncertainty: 0.906%
  • Monte Carlo Simulation Uncertainty: 1.209%

The MCS result is approximately 33% higher than the traditional estimate. This discrepancy confirms that the first-order Taylor expansion used in the sensitivity coefficient method fails to capture the full impact of non-linearities in the thrust model. By underestimating the uncertainty, traditional methods may lead to overconfidence in test data.

Furthermore, MCS proved superior in handling non-Gaussian error distributions. While sensors often follow a normal distribution, other variables—such as pipe diameters or theoretical constants—are better represented by rectangular (uniform) distributions. MCS integrates these diverse distributions more seamlessly than derivative-based methods.

Conclusion

The study highlights that uncertainty assessment methods must be carefully selected based on the mathematical characteristics of the measurement model. For complex, non-linear systems like jet engines, Monte Carlo Simulation offers a more accurate and robust alternative to traditional linear approximations. While computationally more intensive, the ability of MCS to uncover "hidden" uncertainty makes it an essential tool for high-precision aerospace testing.

Summary updated

Abstract

In this paper, Monte Carlo Simulation (MCS) method was used as an uncertainty assessment tool for air flow, net thrust measurement. Uuncertainty sources of the net thrust measurement were analyzed, and the probability distribution characteristics of each source were discussed. Detailed MCS methodology was described including the effect of the number of simulation. Compared to the conventional sensitivity coefficient method, the MCS method has advantage in the uncertainty assessment. The MCS is comparatively simple, convenient and accurate, especially for complex or nonlinear measurement modeling equations. The uncertainty assessment result by MCS was compared with that of the conventional sensitivity coefficient method, and each method gave different result. The uncertainties in the net thrust measurement by the MCS and the conventional sensitivity coefficient method were 0.906% and 1.209%, respectively. It was concluded that the first order Taylor expansion in the conventional sensitivity coefficient method and the nonlinearity of model equation caused the difference. It was noted that the uncertainty assessment method should be selected carefully according to the mathematical characteristics of the model equation of the measurement.
BibTeX
@article{lee2007uncertainty,
  author = {Bo-Hwa Lee and Kyungjae Lee and Inyoung Yang and Soo-Seok Yang and Daesung Lee},
  title = {{Uncertainty Assessment using Monte Carlo Simulation in Net Thrust Measurement at AETF}},
  journal = {International Journal of Aeronautical and Space Sciences},
  year = {2007},
  doi = {10.5139/ijass.2007.8.2.126},
}
Figure