Mech-Health: A Machine Learning Based Fault Detection Using Predictive Analysis For LSTM

Predictive maintenance is a form of maintenance that looks at how well machinery works and how efficient it is during regular operations to reduce the risk of failure. Machinery and other appliances are subjected to less risk due to a sudden failure with the help of predictive maintenance. This form of maintenance device alerts the owner until a malfunction or breakdown is anticipated. Predictive maintenance schedule operations, unlike traditional preventive maintenance, are focused on data obtained from sensors and analysis algorithms. We address the validity of using LSTM instead of conventional approaches such as visual inspection in this paper. Two different models, the Prophet and the Autoencoder LSTM recurrent neural network are used to train the ball-bearing dataset. The prophet method is a forecasting technique implemented with the aid of an encoder-decoder LSTM architecture in R and Python, while Autoencoder LSTM is an autoencoder of sequence data implemented with the help of an encoder-decoder LSTM architecture. Using the two algorithms, and efficient maintenance scheme can be created to limit any potential risk to the assets until they misfire and fully collapse.