Prognotron™ Surface Defect Analyzer

  • Home
  • Prognotron™ Surface Defect Analyzer

Prognotron™

Industrial AI Platform for Smart Manufacturing

 

AI-Powered Visual Quality Control and Defect Assessment for Manufacturing Industries

Product Overview

Manual quality inspection of manufactured products can be repetitive, subjective, and highly prone to human assessment errors. Our state-of-the-art Industrial AI engine enables industries to build and deploy vision and asset analytics AI applications faster, drastically reducing the conceptualization to solution deployment timeline from few months to just few weeks. It comprises of end-to-end Industrial AI pipeline and integrated tools for data acquisition, preparation, labelling and management, secure API development and deployment. The Industrial AI engine learns from historical data and helps to automate the inspection process, removes subjectivity, and eliminates inspection errors. The AI engine can identify defective products, locate product defects (e.g. surface defects, foreign particles, assembly errors, packaging errors), accurately classify defect types and severity levels. It then provides visual feedbacks to the operator or feedback control signals to the automation systems. It can be integrated with wide range of vision hardware systems, and can be deployed on-premise servers, private cloud servers. 

We are looking forward to collaborating with B2B customers including industrial automation Original Equipment Manufacturers (OEMs), System Integrators and Companies in manufacturing sector (e.g. food and beverages, semiconductor, biopharmaceutical and automotive).

Technology Features and Advantages

The Industrial AI engine leverages proprietary state-of-the-art AI technologies to achieve high level of inspection accuracy with a minimal amount of training data, and incrementally learns and improves over time.

 Hardware-agnostic

  • Supports wide range of vision hardware systems that adhere to our data quality guidelines
  • Can be integrated with existing vision systems on production lines and enhance the vision inspection performance using AI engine

AI for High-resolution Data with Fast Inference Speed 

  • AI engine processes high-resolution images in real-time
  • Prevent loss of critical information by preserving pixel-per-defect ratio 
  • High speed, low latency AI Engine – Tested on Private/Public Cloud, On-Premise servers with GPU

Continuous ML/DL Performance Auditing and Learning  

  • Optimized ML/DL pipeline with continuous self-monitoring and assessment to deliver consistent and improved performance over time 

Optimized AI Packaging and Integration

  • Easy integration with existing solutions and OEM softwares through API endpoints hosted on remote or on-premise servers 

Customer Benefits

  1. Cognitive Automation – Eliminate errors due to subjectivity and cognitive load in identifying and assessing defects
  2. Highly Efficient – Automated detection and assessment of various defect types within seconds
  3. Quality Assurance – Ensure quality, safety and compliance of manufactured products and incoming products
  4. Industry 4.0 – Centralized data monitoring of the entire production and procurement quality
  5. Traceable – Transparency and traceability of product quality through digitalized records
  6. Non-invasive – Smart inspection without damaging the products
  7. Mobile – Cloud-based web UI that can be operated on a mobile device and anywhere in the manufacturing plant to quickly record and assess defects on products.

Applications

  1. In-line automated product inspection and quality control on production lines
  2. Quality control of incoming parts and components at procurement stage
  3. Mobile inspection of products and parts (operated on mobile devices like iPad)
  4. Enhancement/Replacement of traditional hard-coded vision inspection solutions with data-driven AI-powered vision inspection

Frequently Asked Questions (FAQ)

Inference speed of AI API (varies based on GPU) 

  • Linux w/ Single NVIDIA Tesla T4 GPU: 55ms per image (defect detection)
  • Linux w/ Single NVIDIA Quadro RTX 4000 GPU: 80ms per image (defect detection)

Supports up to 10 parts per second or 600 parts per minute

  • Supports color and monochrome image formats
  • Tested up to 4K image resolution (12MP)
  • Supports higher resolution (> 12MP) images (with trade-off to inference speed)
  • Frame rate: Up to 10 FPS for defect detection, Supports up to 30FPS for object detection

A platform customizable to industry specific challenges!

Talk to us and find out how Prognotron™ can address your pain points.