Learn · PdM Program Guide + 9 topics
Learn industrial AI for asset maintenance
Start with the PdM Program Guide if you are setting up a program. The nine machine-learning topics then each start with the problem, shows the classical baseline in one short box, then explains the machine-learning approach with interactive illustrations, key takeaways and a short quiz.
Start here · PdM Program Guide
Set up a predictive maintenance program
An easy, end-to-end guide for program managers and reliability engineers: what to monitor, how, how often, and how to prove it works.
Roadmap: setting up a PdM program
The five parts of a program, a ten-step roadmap, and why programs stall after the pilot.
Strategies and the P-F curve
Reactive to proactive, the six failure patterns, P-F intervals by technique and how to set measurement intervals.
Asset selection and criticality
Which assets suit PdM, the asset register, consequence × likelihood scoring and what each tier gets.
Failure modes and techniques
Vibration, ultrasound, motor current, oil, thermography and process data, matched by failure mode, asset type and speed.
Vibration sensors, mounting and settings
MEMS vs piezo, mounting and frequency response, placement, Fmax, lines and record length.
IIoT and wireless monitoring
Architecture, wireless networks, battery life, edge analytics, cybersecurity and hazardous areas.
Rollout, alarms and KPIs
Pilot, baseline and alarm layers, alert-to-work-order workflow, skills, KPIs and a first-pass ROI.
PdM Program Setup
Apply the guide to your own asset list and export a monitoring plan with sensor counts and settings.
Machine learning · Foundations
Why and how
Why AI for asset maintenance
Fixed alarms vs context-aware models, the P-F curve, and when ML is the wrong tool.
The AI-driven PHM pipeline
Sensors, edge, models, decisions and the feedback loop that keeps models honest.
Data and feature engineering
Statistical, spectral, envelope and learned features, explored on simulated faults.
Detection · Diagnosis · Prognosis
The three PHM questions
Anomaly detection
Isolation Forest, autoencoders and forecasting residuals. Learn normal, flag the rest.
Fault classification
Decision boundaries, model choice, confusion matrices and honest evaluation.
Remaining useful life
Health indicators, degradation models, survival analysis and uncertainty that narrows.
Beyond the basics
Images, messy data and production
Computer vision for maintenance
Spectrograms as images, CNNs, transfer learning, thermal and visual inspection.
Real-world data challenges
Imbalance and the accuracy paradox, scarce labels, synthetic data and domain shift.
Advanced and emerging
Explainable AI, physics-informed ML, digital twins, edge AI, drift and LLM copilots.