Share
Export Citation
Wind Tunnel Data Classification A Comparison of Performance Between K-Nearest Neighbor and Artificial Neural Network Algorithm
Purwadi P.
International Conference on Electrical Engineering Computer Science and Informatics Eecsi
Abstract
Wind tunnel experiments provide valuable insights into the aerodynamic characteristics of objects and play a vital role in various engineering and scientific applications. Due to this reason then the quality of the data resulted from the wind tunnel measurement must be maintained in good performance by means the data must be kept accurate and valid. This research is conducting in Indonesian Low Speed Tunnel (ILST), focuses on the classification of wind tunnel data by employing machine learning algorithm. The data that analyzed in this study are pressure measurements, including the total pressure (Pt3), static pressure (Ps4), and the differential static pressures between each side of wind tunnel test section named dP14, dP24 and dP34. The dataset is explored from the ILST database related to the pressure measurements as mentioned above. The dataset was preprocessed to remove noise, handle missing values, and normalize the features. Subsequently, k-Nearest Neighbor (k-NN) and Artificial Neural Network (ANN) model were designed and trained using the preprocessed data to classify different flow patterns based on the pressure measurements. The performance of the algorithms model such as accuracy, precision, recall, and F1-score were evaluated. Additionally, cross-validation techniques were employed to assess the generalization ability of the model. The experimental results demonstrated that the ANN algorithm achieved high accuracy more than 95% when the epoch is selected at the number of 3000, showcasing its potential as an effective tool for data health recognition and faulty analysis. Overall, this research showcases the potential of Artificial Neural Networks as a powerful tool for wind tunnel data classification, enabling more comprehensive analysis to mapping and finding faulty of measurement data of ILST.