PumpSpectra: An MCSA-Based Platform for Fault Detection in Centrifugal Pump Systems

Area of Science:

• Industrial engineering
• Mechanical engineering
• Data analytics

Background:

• Detecting centrifugal pump faults in industrial settings is difficult due to harsh conditions and limited sensor access.
• Fast, explainable fault decisions are crucial for industrial operations.

Purpose of the Study:

• To develop and validate PumpSpectra, an industrial Motor Current Signature Analysis (MCSA) platform for detecting mechanical faults in centrifugal pumps.
• To assess the accuracy, efficiency, and explainability of the PumpSpectra platform in a real-world industrial environment.

Main Methods:

• Developed PumpSpectra, an MCSA platform processing stator-current CSV files using Fast Fourier Transform (FFT) and Short-Time Fourier Transform (STFT).
• Implemented transparent, rule-based models for identifying mechanical faults like misalignment, bearing defects, and impeller anomalies.
• Conducted field validation at a desalination plant, analyzing 40 operating points.

Main Results:

• PumpSpectra achieved 91.2% diagnostic accuracy in identifying pump faults.
• Analysis time was reduced by 95% compared to manual MCSA post-processing.
• A low false-positive rate of 3.8% was recorded at 0.1 Hz resolution, with successful detection of misalignment.

Conclusions:

• Current-only, explainable analytics can effectively support predictive maintenance programs.
• PumpSpectra accelerates fault triage, enhances decision traceability, and reduces maintenance costs in pump-driven industrial assets.
• The platform demonstrates the viability of advanced data analytics for improving the reliability of critical industrial equipment.