Non-Linear Regression Models with Vibration Amplitude Optimization Algorithms in a Microturbine.

Area of Science:

• Mechanical Engineering
• Computational Intelligence
• Condition Monitoring

Background:

• Excessive vibration in rotating machinery can lead to system damage and failure.
• Accurate modeling of system dynamics is crucial for preventing vibration-related issues.
• Non-linear model parameter fitting presents a significant challenge in mechanical systems.

Purpose of the Study:

• To apply the Gray Wolf Optimizer (GWO) algorithm for fitting non-linear regression models to vibration data.
• To evaluate the effectiveness of GWO in optimizing vibration amplitude models for gas microturbines.
• To assess the predictive accuracy of various non-linear models using statistical criteria.

Main Methods:

• Utilized the Gray Wolf Optimizer (GWO) algorithm, an evolutionary computation technique.
• Fitted non-linear regression models (2nd and 4th order, Gaussian, sinusoidal) to radial vibration amplitude data.
• Employed Root Mean Square Error (RMSE) and coefficient of determination (R2) as evaluation metrics.

Main Results:

• GWO successfully optimized non-linear models for vibration amplitude in relation to rotational frequency.
• Improved model fits were observed for 2nd and 4th-order, Gaussian, and sinusoidal models.
• All evaluated models demonstrated high predictive accuracy, with R2 values ranging from 85-93% and RMSE between 0.19-0.22.

Conclusions:

• The GWO algorithm is a viable and effective tool for optimizing non-linear regression models in mechanical vibration analysis.
• The proposed methodology enhances machinery condition monitoring and failure analysis through accurate predictive modeling.
• Statistical analysis confirmed the good agreement and predictive power of the optimized models.