Predictive Maintenance for Gas Distribution Networks
Engagement Context
The client is a leading gas distribution network responsible for supplying piped natural gas to various sectors.
Goals
The primary objective was to create a robust valve and pump health monitoring system. The client needed to enhance predictive maintenance capabilities, save operating expenses, and decrease Non-Productive Time (NPT), all of which were critical to cutting down on unanticipated downtime and system failures.
Solution Approach
The deployed solution incorporated a dual AI strategy, utilizing both supervised and unsupervised approaches to monitor equipment health.
Supervised Approach
The system distinguished between healthy and unhealthy valve actuators by analyzing sensor data that measured hydraulic pressures and flows during valve opening and closing events.
Unsupervised Approach
To catch unexpected anomalies, the system detected sudden, abrupt changes in valve behavior by continuously comparing sensor readings from consecutive opening or closing events.
Value Delivered
The final solution accurately predicted failures in advance, equipping the client with the precise data needed to take proactive maintenance action. The AI anomaly detection systems were successfully trained to make automated, instant decisions based on the underlying circumstances of the equipment.
Ultimately, the solution helped the client to:
- Reduce downtime
- Increase productivity
- Improve safety
- Prevent leaks