Client is a global automotive supplier with the mission to deliver new technology and mobility solutions. With an international team focused on delivering superior value, they have manufacturing facilities across the globe and committed towards delivering top quality products and superior value through innovative processes as well as World-class manufacturing.

The Problem

Like many automotive suppliers, this client was struggling to streamline their manufacturing process to identify part defects in their line. Defect identification often involved manual inspection which was time consuming and had high chances of human error- several of the part “splits” were quite small. This client needed an automated solution to streamline their defect detection and reduce manual efforts.

  • Many parts were having splits

  • Manual inspection difficult to find small splits

  • High chances of human error

  • Time consuming for affects the cycle time

  • Need of Computer vision based solution to reduce the cycle time and improve the quality check process

The Solution

DT40 implemented a Vision Quality Inspection solution to reduce cycle time and improve quality control. 40 cameras were installed to continuously collect data, monitor each part, and identify any possible defects. The cameras were configured to capture frames when a part was visible on the process belt- these images were used to build a deep learning model which was used to classify and identify split defects. DT4o utilized OT data to build a predictive quality machine learning model and output a computer vision based inference system.

The customer experienced 40% reduction in scrap and 47% increase in yield post deployment of this solution. The ability to scale rapidly, and advanced analytics capability this solution offers enables customer’s continuous improvement objectives.

Projected Metrics

The Metrics

  • Throughput of quality control increased by 3X

  • 99.8% accurate inspection models

  • Transition from manual inspection to 100% AI powered vision-based detection

  • Real-time edge performance with 30%+ ROI

The Impact

  • Performance improvement metrics

  • Zero defects

  • Reduce their CoQ by 5% to 30%

  • Improve customer satisfaction

  • Scalability across multiple plants

Vision Quality Inspection App

DT4o’s Vision Quality Inspection App from our AI/ML platform, leverages machine vision data and provides meaningful insights and enables manufacturers to lower cost of quality by reducing scrap, rework, and warranty cost.

At its core, DT4o’s Intelligent Apps are AI/ML powered operational and energy insights solutions that provide Edge data connectivity with an invariant asset hierarchy definition between the edge and the cloud. Through this solution, DT4o builds a robust data pipeline from the Edge to a VPC for telemetry data bolstered by security and best practices pre-configured from AWS.

Capabilities

  • Defect Identification: Detecting quality issues: good part, bad part

  • Defect Classification : Weld missing detection, Too thick or thin detection, Porosity detection. Split detection, and Tear detection

  • Reduction in Cost of Quality: By reducing scrap, rework, customer escapes, returns and warranty costs, customers can reduce their CoQ by 5% to 30%.

  • Increase in Throughput: Automated quality control processes, enhances manufacturing throughput.

  • Accuracy :Advanced CV algorithms from DT4o enables manufacturers achieve near perfect precision levels in production and quality control, with 99.9999% accurate inspection models

  • Enhances Customer Satisfaction: Accelerates customer’s journey to zero defects, and this improves customer satisfaction.

  • Energy costs saving by 10 – 30%

  • Serialize part number: Documenting quality issues in a digital representation.

  • Predictive Quality: With Insight OT Data + Machine vision Quality data facilitates predictive quality at machine/line/equipment level.

AWS Services Used

AWS IoT Core, Amazon S3, Amazon Sagemaker, AWS Kinesis