kev107034011/pdm-lstm-prediction

#Bidirectional LSTM #Optuna #PdM

Predictive Maintenance(PdM) using Bidirectional-LSTM with Optuna

Dataset Description

Maintenance strategy can mainly be categorized into 3 strategies, Run-to-Failure (R2F), Preventive maintenance (PvM), and Predictive maintenance (PdM).

R2F: Only happens when equipment stops working PvM: Time-based maintenance/ Scheduled maintenance: periodically maintain usually based on time or process iteration PdM: Use predictive tools to determine when maintenance action needs to be taken

The goal is to predict the Remaining Useful Life(RUL) of each type of machines.

All data comes from C-MAPSS Aircraft Engine Simulator Data with noise features added.

Data Preprocess

Mainly referenced the preprocess methods from umbertogriffo and PHAM VAN VUNG

Algorithms

• Stacked Multiple Bidirectional LSTM layers with Attention layer to build the prediction model.
• Utilized Optuna with TPESampler to automatically optimize hyperparameters, including Learning Rate, Number of Layers, Nodes, Dropout Rates,etc.

Code

Bi-LSTM Model with Optuna