DamTurbineVibration&AcousticMonitoringGrid
High-frequency piezoelectric telemetry and localized Fourier FFT inferencing across twelve 37.5MW Kaplan water turbines.
The Brief
Keystone River Power Authority operates a 450-megawatt hydroelectric dam facility containing twelve vertical-shaft Kaplan turbines commissioned between 1954 and 1978. Unscheduled turbine shutdowns caused by cavitation erosion, bearing misalignment, and shaft wobbling cost up to $180,000 per day in replacement power purchases on regional wholesale energy markets.
Inspections required draining turbine draft tubes and sending technicians into confined water passages with ultrasonic thickness gaugesa hazardous procedure that could only take place twice per year during low river discharge seasons.
Apache Spark was selected to design, wire, and deploy a continuous acoustic and vibration diagnostic monitoring grid capable of detecting microscopic mechanical fatigue months before operational failure occurred.
Engineering Approach
We engineered and installed an array of hermetically-sealed triaxial piezoelectric accelerometers and acoustic emission sensors directly onto turbine guide bearings, shaft seals, and generator stator frames. Sensor signals are carried through shielded mineral-insulated cable to ruggedized data acquisition (DAQ) units deployed inside turbine gallery galleries. Edge compute nodes continuously compute Fast Fourier Transform (FFT) harmonic spectra, comparing live harmonics against baseline baseline mechanical signatures to identify sub-millimeter cavitation pitting and bearing race spalling.
Operational Outcome
The monitoring system detected micro-fracture propagation in Turbine 7s lower guide bearing 11 weeks before catastrophic mechanical failure, enabling scheduled repair during a planned diversion window and saving an estimated $14M in lost generation and equipment damage.
Schematic Delivery Pipeline
Submersible triaxial accelerometers capture high-frequency bearing vibrations at 25.6 kHz.
Low-noise pre-amplifiers digitize micro-volt sensor signals inside concrete dam galleries.
Edge compute algorithms decompose raw waveforms into harmonic frequency spectra.
Automated deviation alarms alert plant operations engineers to anomalous harmonic spikes.
Legacy vs Modernized Architecture
Technical Plates & Trace Graphs
Kaplan Turbine 07 lower guide bearing accelerometer mount and conduit routing.
Harmonic vibration waterfall chart showing early micro-fracture frequency spike.
Turbine inspection gallery DAQ panel and optical ring backbone termination.
“Apache Spark saved this utility fourteen million dollars on Turbine 7 alone. Their diagnostic system detected bearing fatigue that would have gone unnoticed until the turbine tore itself apart.”