Five layers of intelligence.
One unified platform.
Each layer builds on the last. Start with the data you already have. Add sensors and ERP when you're ready. Every layer makes the next one smarter.
Reactive Maintenance Intelligence
Understand what's already broken — and why.
Most factories know their machines break. They don't know which machines break most, which faults cost the most, which technicians resolve fastest, or which repairs keep repeating. Flarehex mines your existing maintenance log and answers all of it automatically.
What you upload:
Any CMMS export or Excel maintenance log. Columns: case ID (fault/work order number), activity (what happened), timestamp (when it happened). That's it.
What Flarehex finds:
- Actual repair paths — not assumed flow, real sequence discovered from data
- Top 5 failure causes ranked by total downtime cost in dollars
- Average time per fault type per machine per technician
- Repeat failures — faults that should have been resolved permanently
- Fastest resolution pattern — the path that closes faults quickest
Sample AI Insight
“Machine 3 accounts for 43% of your total downtime. The average repair takes 6.2 hours. Technician A resolves this fault type in 3.1 hours — less than half the average. Assigning her by default to Machine 3 faults would save approximately 18 hours of downtime per month.”
Visualisation: Discovered Process Map
BPMN-standard process maps rendered in real time from event logs.
Visualisation: Conformance Overlay
Planned vs actual process comparison side-by-side.
Planned
Actual (Night Shift)
Preventive Maintenance Compliance
Are your scheduled PMs actually happening — or just logged as happening?
Production pressure is the enemy of preventive maintenance. Operators skip steps. Night shifts cut corners. Managers have no visibility. Flarehex's conformance checker compares your planned PM schedule against what the data proves actually occurred — and shows you every deviation, automatically.
What you add:
Your PM schedule (planned tasks + intervals) alongside your existing maintenance log from Layer 1.
What Flarehex finds:
- Skipped steps — tasks planned but never executed
- Wrong sequence — steps completed out of order
- Late completions — PMs done days or weeks after due date
- Shift-based deviation — which shift has lowest compliance rate
- Correlation — which skipped PMs directly preceded equipment failures
Sample AI Insight
“Lubrication check on Machine 5 is skipped 67% of the time on night shift. Cross-referencing with Layer 1 data: bearing failures on Machine 5 occur 3.4x more frequently in the week following a skipped lubrication. Estimated cost of this single deviation: $4,200/month in reactive repairs.”
Predictive Maintenance
Know a machine will fail before it does — not after.
Reactive and preventive maintenance still leave gaps. Unexpected failures happen even with perfect PM compliance. Layer 3 adds real-time condition monitoring via IoT sensors and NVIDIA-powered edge inference — so you know a machine is degrading hours or days before it stops.
What you add:
An NVIDIA Jetson Orin Nano edge device (~$500, purchased once by you) plus off-shelf vibration, temperature, and current sensors attached to critical machines. AWS Greengrass manages the device remotely — no on-site IT visits after initial setup.
What Flarehex finds:
- Vibration anomalies: deviation from baseline pattern (σ-score)
- Thermal trending: gradual temperature rise indicating friction or wear
- Current draw changes: motor degradation signature
- Composite anomaly score: multi-sensor failure probability
- Estimated time to failure: confidence-interval forecast
Sample AI Insight
“Machine 3 bearing vibration is 2.3 standard deviations above its 30-day baseline and rising at 0.4σ per day. At current rate, estimated failure in 48–72 hours. Confidence: 84%. Recommended action: schedule repair for tomorrow morning before shift 2 begins.”
Visualisation: Sensor Dashboard
Live vibration, temp, and current tracking with anomaly timeline.
Vibration
2.3 sigma
Temp
78.4 C
Current
12.1 A
Anomaly Timeline (last 60 min)
Visualisation: Revenue-at-Risk
Cross-domain causality graph linking maintenance to production orders.
Affected Orders
PO-1842, PO-1851
Revenue at Risk
$12,400 / 48h
Process Intelligence
The machine that's about to fail — which customer orders does it put at risk?
This is the question no other tool answers. Maintenance systems know about machines. ERP systems know about orders. Nobody connects them. Layer 4 is the first tool at SMB price to build a unified process map that spans from sensor anomaly to production delay to customer delivery risk — and quantifies it in dollars.
What you add:
Your ERP production log (work orders, production orders, delivery dates). A standard CSV export from Odoo, SAP B1, or any ERP. Combined with Layers 1–3, Flarehex builds a unified cross-domain event log.
What Flarehex finds:
- Which machines are single points of failure for current orders
- Which orders are at delivery risk given current machine health
- Revenue exposure per asset per week
- Upstream cause of every production delay (traced back to maintenance event)
- True cost of each failure: repair time + production delay + customer impact combined
Sample AI Insight
“Machine 3 shows 84% failure probability within 48 hours. It is the only asset certified to produce Part #A47. Two open orders depend on Part #A47: Order #4471 (Acme Corp, $28,000, due Friday) and Order #4502 (Beta Corp, $11,000, due Monday). Scheduling repair for Tuesday morning eliminates $39,000 in delivery risk. Delaying repair risks both orders and a potential late-delivery penalty of $5,800.”
Resource Optimisation
Given everything you now know — here is the optimal plan for the next 7 days.
Layers 1–4 give you the full picture. Layer 5 acts on it. Using Google OR-Tools constraint solver and the unified event log already built, Flarehex generates an optimised schedule for technicians, machines, materials, and energy — simultaneously, updated in real time as conditions change.
What it optimises:
- TechniciansWho to assign to each fault, based on skill match, resolution speed history, current workload, and failure priority.
- MachinesWhich machines to load heavily vs rest. Avoids scheduling high-load runs on assets within a predicted failure window.
- Spare parts & materialsPre-orders replacement parts before predicted failures based on supplier lead time.
- EnergySchedules high-draw processes to off-peak utility tariff windows. Typical saving: 8–15% on monthly energy bill.
- Shift rosterOptimal crew allocation across shifts given production targets and constraints.
Sample AI Output
“Optimal 7-day schedule: Assign Tech A to Machine 3 Tuesday 08:00 (pre-failure repair, 3h window). Load Machine 5 to 90% Wednesday–Friday (healthy window, no predicted risk). Pre-order bearing #B224 today (3-day lead time, needed by Thursday). Move press run to Sunday 23:00–03:00 (off-peak rate saves $340 this week). Estimated output impact: zero disruption to all 12 open orders.”
Visualisation: Optimised Gantt
7-day resource schedule combining technicians, machines, and energy.