envisma

Process Vision

Built around your process, not ours

Every Process Vision deployment is configured from a digital work instruction, developed with your engineering team. It defines each step of the build, so the system does not simply hunt for anomalies - it knows what should be happening, and tells you when it isn’t.

Digital work instruction: an assembly network of build steps with parallel branches, positions, tolerances and zones, mapped onto a blueprint of the blade mould with nominal ply outlines and a critical zone marked

Key point 1 - Configuration

The instruction encodes the process you already run

Adopting Process Vision does not mean changing how your factory works. Configured in a few days, the digital work instruction can be reused across blade variants.

MaterialsReinforcement types, fabric, unidirectional, preformed stacks, etc.
SequenceEvery process step, in order, with sequence numbers
Positions & tolerancesMeasurement against nominals for sub-elements, consumables, equipment
Zones & severityWhich areas matter most, and what counts as critical

Key point 2 - The record

Nothing is sampled. Nothing is reconstructed.

Every image captured on the mould is analysed and kept - annotated frame by frame with what the AI detected. Not a folder of footage, and not a summary written afterwards: the build itself, on the record.

Annotated mould camera frame with AI detections: a wrinkle and a foreign object boxed and labelled on the live camera scene
Every valuable frame retainedThe full image set for every component produced
Every detection annotatedActivities, materials, personnel, anomalies, defects
Event-driven traceabilityReports synchronised with factory equipment operation

Key point 3 - Control

Inspection tells you what went wrong. Process Vision changes what happens next.

Real-time detection is only useful if something happens as a result. Process Vision closes the loop back to the equipment and the people on the floor.

Factory equipmentDirect integration with PLCs and a REST API for system and factory control
Infusion gate controlMulti-gate resin infusion monitored and controlled, with flow simulation
Operator guidanceLaser projection for instruction and warnings, visual alerts to mobile devices
Management reportingDashboards, ERP and data warehouse integration, factory-to-factory comparison

Process control - layup

Ply Identification & Tracking (PIT)

PIT automates the layup sequence so operators spend their time placing plies rather than working out what comes next. An "Assembly Network" containing all possible ply layup sequences is built. The system then tracks the layup sequence, automatically advancing the laser projection of plies as the layup progresses.

Verified as it is placedAI detection reads the ply index to confirm it is the right ply, and checks its position against tolerance to confirm it is in the right place. Approval is only ever given when a ply is correct, in sequence and in tolerance.
A snapshot per placementWhen a placement is approved, the laser outlines advance and a quality snapshot is saved as evidence - available on your internal network the moment it is taken, and retained against the blade long after it leaves the mould.
The mould stays clearBecause verification happens as the ply is placed, fewer quality technicians are needed on the mould. The tablet doubles as a layup work instruction that is always current and always to hand.

Carbon reinforcement placement and adhesive deposition work in a similar way: the nominal centrelines are laser-projected onto the mould surface, and carbon or adhesive is verified against the work instruction as it is placed. Controlling layup with Process Vision brings the added benefit of a complete, certifiable quality record produced as part of the work itself.

Paired blade mould shells with ply verification overlays: identified plies outlined and labelled, and a wrinkle flagged on ply 58

Envisma partners with LAP to integrate Process Vision and LAP's PRO-SOFT, enabling seamless laser-projected feedback directly onto the mould surface.

Process tracking

Takt time, timelines and timelapse - with no manual data entry

Blade-type-specific models detect each process step from mould preparation through to demould, so the pace of the line is measured by the system, not written down by hand.

Takt time monitoringEvery step is timed automatically and compared against your target takt. Track it by site, production line, shell or step to see where your pace is really being set - and whether you are on course for annual output.
TimelinesThe same AI detection maps every cycle as a Gantt chart. Split timings by line, shift team, day of the week or time of day, and the pattern behind your slower cycles becomes obvious.
Intelligent timelapseMulti-camera arrays generate massive, redundant streams; the system condenses them into high-resolution visual timelapses, browsable by timestamp and timeline, with a 60-day image history available to users through their web browser.
Activity timeline: detected process steps for four blade builds shown as a Gantt chart, with build times per blade and a cure step flagged over takt

Industry focus - wind turbine blades

Detection built for blade moulding

Deployed on large-scale wind turbine blade moulding operations worldwide, with detection capabilities actively developed for the industry.

Wrinkles in reinforcement layup

Position, size, orientation and height
Severity assessed against quality measures
Prioritisation by specified zone

Resin infusion monitoring

Flow front propagation, position, speed, acceleration
Leaks, excess resin, aeration, air entrapment
Watch-location fill synchronised with manifolds

Sub-element positioning

Location of equipment, sub-elements, prefabricated parts
Alignment relative to nominals, e.g. web

Sequence validation

Material type, orientation, sequence number verification
Reinforcement classification: fabric, unidirectional, chopped
Foreign objects during layup or infusion
Paired blade mould shells viewed root to tip, with a wrinkle, a foreign object and an anomaly boxed and labelled across five camera views

Platform

What the system does, end to end

Visual data acquisition
2D and 3D calibration in real-world coordinatesFusion with thermal, temperature, pressure and humidity sensorsPolarisation, on-camera compression, automatic lens cleaning check
Detection & classification
Activities during manufacturing operationsAnomalies and obstructions, e.g. gantry cranes, peopleData augmentation from factory PLCs
User interfaces
Web application for stream view, process control, analysis and reportingTimelines, 2D/3D and augmented reality visualisationQuality record per component, unified defect logging
Network & equipment
Integration with factory equipment, e.g. PLCsLaser projection for auto-calibration, instruction, alertsData warehousing, ERP integration, REST API

From images to insight

Big data analysis across time, space and sites

Time per step, per operator, per mouldWorkforce activity and operational time broken down by process step, with activity heatmapping over time.
Comparison across factories and partsSite against site, and component against component in serial production.
Into your existing stackHigh-level management reporting integrated with company ERP and data warehouses, plus REST API export.

Partnership

Resin infusion control with LMAT

A Gate Control system for monitoring and controlling multi-gate resin infusion - sensors, resin flow simulation, customised PLC systems and the related user interfaces, connected directly to Process Vision.

01Process Vision detects flow front and anomalies
02Simulation predicts fill across the mould
03PLC adjusts gates during infusion
04Event report closes the loop, fully traceable
Blade mould viewed root to tip during resin infusion: the infused region along the centreline shown darker, outlined in magenta by the detected flow front, with gate states marked

Ease of deployment

Configured and managed with local factories

40+ installations running across Europe, North America, China and India. Camera layouts are planned against a digital twin of your factory, and the system integrates with laser projection systems already on your floor.

EuropeNorth AmericaChinaIndia
Digital twin factoryCamera layouts planned and validated before installation
Laser projectionIntegration and control with various laser projection systems
Local configurationDeployments configured and managed with local factories

FAQ

The questions every factory asks first

Where does our production data live?
All AI runs locally, on servers in your factory. Production images and quality data never leave your site - no cloud dependency, real-time performance at production speed, and simpler IT security approval.
Do you train a shared model on our images?
No. Nothing is pooled, nothing is shared. Where images come to our servers to train your models, they stay walled off and are used for one purpose only: your system. The models trained on your blades are yours alone - not a tier of service, but how Envisma is built.
How fast is a deployment configured?
The digital work instruction is configured with your engineering team in days, and reused across blade variants.
What does the system integrate with?
Factory PLCs, laser projection systems, ERP and data warehouses - plus a REST API for data export and factory control.
What happens when the internet goes down?
Nothing. Detection and alerts run on-premises, immune to internet outages and latency.

See it against the defects you actually care about.

Book a meeting How the trial works