AI to Smarten Up Automated Optical Inspection
In the Bosch vision, the factory of the future will feature the ability to reconfigure itself to changing conditions such as product variance, production fluctuations and process adaptations – a new level of flexibility compared to today. That requires systems for automated optical inspection (AOI) to evolve from product-specific systems to task-flexible and convertible solutions. The Bosch visual inspection assistant (ViPAS) is the first step towards a redefined AOI-system.
In industrial manufacturing, product quality control is essential in order to fulfill the highest quality demands of customers. Automated production lines accomplish this in part with highly customized optical inspection systems. Such systems usually comprise cameras that capture images of parts to be checked and a task-specific software for automatically detecting defects on those products. Automated inspection is nothing new, but in a digital age, the flexibility of automation is reaching new heights.
A New Challenge
Today’s customized industrial vision inspection systems were a big step forward in fully automated inspection over the past decades. Now, a key aspect of Industry 4.0 is making production more flexible, and the need for task-flexible automated optical inspection (AOI) has arisen.
In the future, smart factories will contain machines that can dynamically react to new production requirements. Production lines will no longer be dedicated to single, high-volume products. They will be adaptable, flexible, convertible and reconfigurable – manufacturing different product variants and generations.
Bosch has developed an advanced automated Visual Inspection System (ViPAS), which is designed to adapt quickly in the factory of the future. Using Deep Learning approaches, ViPAS will be a key technology in Industry 4.0.
Automated optical inspection (AOI)
Consequently, automated optical inspection systems have to become adaptable, task-flexible and reconfigurable. These advancements – aligned with the flexible nature of smart factories – are made possible with artificial intelligence (AI). With this vision in mind, Bosch Research has developed the Visual Inspection Assistant, or ViPAS, a prototype of a new product-flexible AOI system. In addition to a robotic arm that can grip any kind of part, the prototype includes multiple state of the art cameras that can capture 2D and 3D images from up to nine perspectives as well as dedicated brightfield and darkfield illumination modules.
Deep Learning approaches provide a key technology for a trainable defect detection system. Specifically, it makes it possible to train and ramp-up ViPAS for a new task simply by presenting sample data to the system. The machine learns on its own after that, eliminating the need for task-specific programming. It then quickly decides if a product is defective or not. Importantly, ViPAS can be configured, trained and operated via a simple graphical user interface with no need for programming expertise.
Deep Learning uses algorithms to analyze data and reach decisions independently the way humans do. It means a machine can learn to recognize a defect on its own without pre-programmed knowledge of it.
The Road Ahead
ViPAS is currently an internal R&D prototype and evaluation system for future visual inspection machines at Bosch. After pilot runs at plants in Nurnberg and Feuerbach in Germany, further evaluations of the concept are planned. Additionally, components of ViPAS are being integrated into existing automation products from Bosch to enhance functionality for customers.
