MAE551_Project

In this comprehensive machine learning project, our primary objective is to tackle the critical issue of predictive maintenance in industrial settings. The project focuses on the development and evaluation of predictive maintenance algorithms utilizing a meticulously designed synthetic dataset, closely mirroring real-world scenarios. With a dataset comprising 10,000 data points and 14 crucial features, encompassing variables such as air and process temperatures, rotational speed, torque, and tool wear, our aim is to establish a robust super learning algorithm. This algorithm integrates Convolutional Neural Networks (CNN), Random Forests, and Gradient Boosting methods as base learners. The overarching goal is to create a model capable of forecasting machine failures or categorizing specific failure types, thereby contributing significantly to the enhancement of machinery reliability and operational efficiency. Furthermore, we explore an alternative approach by developing a distinct Long Short-Term Memory-Generative Adversarial Network (LSTM-GAN) method. This method is then systematically compared with the super learning algorithm on various performance parameters, providing a comprehensive analysis of their respective strengths and weaknesses. The insights gained from this comparative evaluation aim to guide future developments in predictive maintenance strategies, offering a valuable contribution to the field of industrial machinery reliability and performance optimization.