Development of Predictive Maintenance System for Haemodialysis Reverse Osmosis Water Purification System

Predictive maintenance utilizes a variety of data analytics and statistical techniques to predict possible device or equipment failures and provide suggestions on maintenance strategy according to the results of predictive analytics. This paper presents the development of an IoT-based predictive mai...

Full description

Saved in:
Bibliographic Details
Published in2022 4th International Conference on Smart Sensors and Application (ICSSA) pp. 23 - 28
Main Authors Bani, Nurul Aini, Noordin, Muhammad Khair, Hidayat, Achmad Alfian, Kamil, Ahmad Safwan Ahmad, Amran, Mohd Effendi, Kasri, Nur Faizal, Muhtazaruddin, Mohd Nabil, Muhammad-Sukki, Firdaus
Format Conference Proceeding
LanguageEnglish
Published IEEE 26.07.2022
Subjects
Cloud computing
healthcare modernization
hemodialysis
Machine learning
medical equipment
predictive maintenance
Predictive models
Programming
Prototypes
Purification
Sensor systems
smart healthcare
Online AccessGet full text

Cover

More Information
Summary:Predictive maintenance utilizes a variety of data analytics and statistical techniques to predict possible device or equipment failures and provide suggestions on maintenance strategy according to the results of predictive analytics. This paper presents the development of an IoT-based predictive maintenance system for the Haemodialysis Reverse Osmosis (RO) Water Purification System on three main categories; the mini prototype of the RO system, the hardware and electronics circuit of the RO system and the machine learning programming and dashboard monitoring of the RO system. The mini prototype of the RO system utilizes three types of sensors which are pressure sensors, conductivity sensors and flow sensors. Using the ESP8266 Arduino module, the system has successfully captured the sensors' signals and transmit the data to the cloud storage. The developed web application interface has managed to view the data from the sensors of the working prototype and display them in a graphical form to be used as input for further analysis. The trained LSTM model used is working perfectly as it managed to detect anomalies in sensors' readings and predict the breakdown of the plant.
DOI:10.1109/ICSSA54161.2022.9870965