Machine Learning-Based Failure Prognostication of flight Equipment utilizing Hybrid Techniques

DOI: https://doie.org/10.0618/Jbse.2024131340

Jalajakshi V , Dr. Myna A N

Keywords:

Random Forest, Neural networks, Dataset, RF algorithm, Regression

Abstract:

The air travel sector has a lot of knowledge and general upkeep data that could be utilized to estimate future behavior and produce useful outcomes. This research seeks to use variable selection and information deletion-based machine learning approach for predicting aeroplane system problems. Over the duration of two years, upkeep and inability information for flight systems and equipment have been gathered, and nine input and one outcome value were painstakingly recognized. To increase the accuracy of malfunction count prognostication in 3 levels, a blended model development model is suggested. Among most efficient and unproductive parameters are identified in the first phase using the RF method of feature selection for feature assessment. To remove noisy or inconsistent data, a modified K-means algorithm and naive bayes is used in the stage 2 . SVM, Decision Trees, linear regression, Random Forest and multilayer perceptron such as neural networks and advanced algorithms, respectively, are used to assess the efficacy of the proposed model development model on the upkeep set of data of the appliances. Additionally, in the third stage the designs are assessed using performance measures like the coefficient of correlation, mean absolute error, and root average square error. The findings show that the hybrid feature extraction model is effective in estimating the device failure rate.

PDF