AI enabled machine vision for manufacturing

Envisma Process Vision

Envisma develops AI enabled machine vision systems for large scale industrial manufacturing with the aim to directly influence operations in the factory, improving efficiency and reducing defects. Our Process Vision software controls multi-camera arrays, monitoring production 24/7, tracking the time required and workforce activity for each step in the manufacturing process. AI detection capabilities in Process Vision trigger special detection layers built to identify and alert to defects as they appear in real time.

  • Multiple high resolution cameras arranged in stitched scenes, distributed across different factories and locations.
  • Multiple custom detection layers identify and locate consumables, materials, equipment, personnel, activities, sub-components, anomalies, defects, etc. 
  • Big data analysis of highly complex interlinked time & space related datasets
  • Visualisation of analyses and of real-time actionable insight through alerts & equipment control
  • High level management reporting integrated with company ERP and data warehouses
  • AI models and detection libraries purpose built for clients


The Envisma Process Vision system including its artificial intelligence-based software acquires images and performs analysis to determine events and anomalies occurring in a production environment. The system is regularly deployed on large scale manufacturing operations in industrial applications. The software is continuously developed to improve accuracy and extend its functionality. The PV system also provides a framework for building powerful machine learning and other models for image analysis. Envisma has designed and developed detection model architectures for a range of applications including resin infusion and utilises proprietary methods and software code for training and evaluation of these models.

Visual data acquisition and management 
  • 2D and 3D calibration of image data for reporting in real world coordinates
  • Dynamic re-configuration of imaging devices to adapt to changing detection requirements, lighting, and other environmental effects
  • Archiving of recordings and providing access to archived data
  • Combining / augmenting image data with other sensors e.g. thermal imaging devices, temperature, pressure, humidity sensors
  • Polarisation to enhance detection or reduce reflections
  • On-camera data compression
  • Automatic lens cleaning and device maintenance
Detection and classification
  • Detection and classification of activities during manufacturing operations
  • Detection and mitigation of anomalies e.g. gantry cranes, people, other obstructions
  • Data augmentation with external input e.g. from factory PLCs
  • Personnel activity and movement
  • Heatmapping of personnel activity over time
  • Personnel operational time breakdown by process step
  • Validation of operational / assembly sequence
  • Material classification e.g. type, orientation, sequence number verification
  • In-operation warnings e.g. overhead cranes, equipment misplacement, personnel in dangerous locations etc.
User interfaces
  • Web application interface to camera streams, analysis and reporting
  • Raw data collection and visualisation on timelines
  • Integration with factory equipment e.g. PLC, projection equipment etc.
  • Direct data export via REST API requests from 3rd party clients
  • 2D/3D visualisation of detections including augmented reality
  • Comparison of different factories
  • Comparison of different components in serial production
  • Kiosk and big screen user interfaces
  • Mobile device integration including headsets, wearables, mobile phones etc. 
  • Progress / status dashboards
  • Event triggered remote image capture and synchronisation
  • Management of records
  • Quality record for each component produced using unified defect logging
  • Alerts to operators through augmented reality, audible devices, visual devices
  • Feedback between operators and the PV system
  • Annotation of components during and after manufacture 
  • Augmentation of ML models and using feedback from factory operations

Industries: Wind Turbine Blade Manufacturing

Process Vision is deployed on large scale wind turbine blade moulding operations in globally. Envisma actively develops the following detection capabilities for the wind turbine blade manufacturing industry:

Wrinkles in reinforcement layup
  • Position, size, orientation, height etc.
  • Severity with respect to structural other quality measures
  • Prioritisation with respect to specified zones
Resin infusion monitoring
  • Resin flow front propagation, position, speed, acceleration during infusion
  • Anomalies e.g. leaks, excess resin, aeration, air entrapment etc.
Sub-element positioning
  • Location of operational equipment, sub-elements, prefabricated parts
  • Location and alignment of sub-elements relative to nominals e.g. spar
Validation of operational / assembly sequence
  • Material type, orientation, sequence number verification
  • Reinforcement material classification e.g. fabric, unidirectional, chopped
  • In-operation warnings
  • Foreign objects during material layup or infusion
Resin infusion control
Envisma has partnered with LMAT for the development of a Gate Control system for monitoring and controlling multi-gate resin infusion systems. The system connects directly to Process Vision and includes sensors, resin flow simulation, customised PLC systems and related user interfaces. Capabilities include:
  • Interactive detection of consumables and infusion pipes prior to infusion
  • Synchronisation of watch location fill monitoring with factory equipment e.g. infusion manifolds 
  • Event driven reports synchronised with factory equipment operation to provide full infusion recording and traceability