Sewerage Bearing Fault Detection
Project Scale & Technical Specifications
Assets Monitored
The system monitors 16 main sewage pumps distributed across two major pumping stations.
The facility processes an average flow of 1.7 million cubic meters per day, with a design capacity of up to 2.45 million cubic meters.
The assets are high-capacity industrial units with rated motor power ranging from 2.25 MW to 2.5 MW.
Deployed Sensors
A grand total of 224 sensors are utilized to capture granular operational health data.
208 Wireless Tri-axial Accelerometers: Providing high-resolution waveform and spectrum data.
16 RPM Sensors: For real-time shaft rotation speed monitoring.
Each pump set is equipped with 14 sensors distributed across the motor, intermediate bearings, and pump body.
Annual Data Collection
The system operates at a sampling frequency of 25,600 Hz.
It captures 306,600 raw data samples annually (based on 840 captures per day).
This contributes approximately 613,200 sensor-hours of operational data per year.
Installation Period
Plug&Sense physical sensor and control station installation required only one day.
This was followed by a 2-to-4-week baseline sampling period for machine learning calibration.
After calibration the system was fully calibrated to the sewerage machinery's unique vibrational fingerprint
The Challenge
The Challenge
The client required an independent, holistic monitoring system capable of detecting abnormalities autonomously. The primary objective was to move away from disparate monitoring tools and reactive "run-to-fail" maintenance models.
The critical challenge was to implement a solution that could prevent unexpected pump failures that would interrupt essential public utility services, while simultaneously overcoming significant industrial broadband noise that typically masks the early signs of machine failure.
The System
The System
01
Broadband Noise Elimination
The solution utilizes a proprietary Signal Averaging Strategy to segment signals by shaft period, effectively removing uncorrelated broadband noise while preserving critical health indicators.
02
Early Fault Detection
The system identifies anomalies up to 6 months in advance, preventing catastrophic breakdowns before they occur.
03
2–10x ROI
By eliminating unnecessary part replacements and reducing emergency repair costs, clients achieve a measurable 2–10x return on investment.
04
96.6% Accuracy
Leveraging a Transformer-Based Autoencoder (TA) and an "Unfair Advantage" library of over 10 million operational hours, the system achieves a state-of-the-art 96.6% anomaly detection accuracy.
New Algorithms and Research Breakthroughs
The project introduced a world-first hybrid anomaly detection pipeline known as META. This included Signal Averaging Strategy, Multimodal Feature Fusion and Transformer-based Autoencoder (TA).
META
Multimodal-feature Extraction & Transformer-based Autoencoder
A total of 224 sensors have been installed to capture granular operational health data:
224
Deployed Sensors
Conclusion
Conclusion
Revolutionizing Infrastructure with the META Pipeline
This project introduced a world-first hybrid anomaly detection pipeline known as META (Multimodal-feature Extraction and Transformer-based Autoencoder).
The "Signal" in the Noise To handle the complex acoustic environment of a sewage plant, XTRA Sensing implemented Multimodal Feature Fusion. This technique simultaneously analyzes Axial, Radial X, and Radial Y vibration signals through Principal Component Analysis (PCA) to create a comprehensive "Machine Health Index," ensuring no fault goes unnoticed.
From Reactive to "Virtual Expert"
The facility has successfully transitioned to a model of 24/7 "Virtual Expert" monitoring. This shift empowers field engineers with real-time intelligence, reducing the stress of managing mission-critical infrastructure.
The Outcome: Sustainability and Reliability
Beyond immediate fault detection, the deployment has driven significant ESG goals. The predictive maintenance capabilities have extended asset life by over 25% and reduced energy consumption by more than 18%, ensuring the facility remains efficient, sustainable, and reliable.