Sensors Data Analysis in Supervisory Control and Data Acquisition (SCADA) Systems to Foresee Failures with an...

Area of Science:

• Industrial Automation
• Predictive Maintenance
• Machine Learning

Background:

• Traditional fault detection methods in industrial systems often struggle with faults of undetermined origin.
• There is a need for enhanced prognostic capabilities to enable preventive and predictive maintenance.
• Improving machined product quality and reducing breakdown times are critical for operational efficiency.

Purpose of the Study:

• To design and implement a Supervisory Control and Data Acquisition (SCADA) system for automatic fault detection.
• To leverage Industrial Internet of Things (IIoT) and machine learning (ML) for enhanced fault prediction.
• To improve machine availability and product quality through advanced diagnostics.

Main Methods:

• Integration of SCADA systems with Industrial Internet of Things (IIoT) technologies.
• Application of various machine learning (ML) techniques for anomaly detection and fault prediction.
• Analysis of diverse data sources and strategic replacement of digital sensors with analog sensors.

Main Results:

• The system demonstrates improved prognostic capacity for detecting faults, even those with undetermined origins.
• An effective anomaly detection algorithm was developed to foresee failures before they trigger alarms.
• Significant improvements in machine availability were observed post-implementation.

Conclusions:

• The novel SCADA system, enhanced by IIoT and ML, successfully achieves automatic fault detection and prediction.
• The system contributes to preventive maintenance, better product quality, and reduced downtime.
• The enhanced prognostic capabilities lead to greater overall machine availability and operational reliability.