Predictive Maintenance using Machine Learning with the Support from Smart Sensors and Supply Chain Management using Blockchain
Sankhapani Neog, Kaushik Das
Abstract
Objectives: The objective of the research is to study the existing predictive maintenance solutions, find more efficient solutions that can help industries improve their efficiency, and study the advantages of decentralized supply chains. The efficiency of predictive maintenance systems can be improved by multiple sensor inputs. Also, a user-defined safe limit value system will improve the efficiency of the predictive maintenance system. Methods: Time-series data is used in our machine-learning-based forecasting operations on Google Collaboratory. Long Short-Term Memory (LSTM) and Prophet models are used in our study for time-series forecasting. Findings: For the predictive maintenance system, multiple sensor inputs will lead to more efficient results. Instead of predicting a single sensor value, we can take the idea from Google\'s weather forecast to predict future assets\' health by taking multiple inputs. Novelty: Predicting maintenance systems lack the feature of a user input for safe or unsafe state. By giving a user input for safe or unsafe state the system will understand the failure. Improvements can be done on predictive maintenance systems by giving multiple sensor inputs. Keywords: Blockchain, Corrosion, Maintenance, Machine Learning, Predictive maintenance