Physical AI & predictive maintenance
Anonymized plant pattern: vision/robotics signals and PdM remaining-useful-life joined to orders and capacity — so failure windows reshape promises before the line stops.
Context
A process-heavy plant had rich Physical AI and CMMS predictive alerts. Customer promise dates still assumed full capacity.
Challenge
- Physical AI and PdM were dashboard orphans.
- Emerging capacity loss never hit MRP or ATP.
- Maintenance and planning fought after the break, not before.
Approach
- Joined MES, PdM, and order graph in the supply chain brain.
- Decision recommendations for plan shift and spare PO with human approval.
- RUL shown as capacity impact, not only a health score.
Outcomes (qualitative)
Capacity impact visible before the failure window. Maintenance and planning shared one decision object.
Related reading
- Predictive maintenance operations — remaining useful life tied to capacity
- Decision intelligence solution — how a recommendation reaches approval
- Case study: global manufacturing — demand planning on the same fabric
- Supply chain AI guide — definitions, use cases and evaluation criteria
Frequently asked questions
What is the difference between predictive maintenance and this approach?
Standard predictive maintenance predicts asset failure. This approach converts predicted remaining useful life into lost capacity hours, pushes that into MRP and available-to-promise, and reshapes the plan before the failure window arrives.
Why were physical AI signals described as dashboard orphans?
Vision, robotics and condition-monitoring alerts were visible to maintenance but never reached planning, so customer promise dates continued to assume full capacity while the plant was degrading.
Who approves the resulting plan changes?
Planners and maintenance leads share one decision object. Recommendations for resequencing and spare-part orders are raised with their capacity impact and wait for human approval.
Download this case or see live PdM ops.