CASE STUDY: Manufacturer of automobile

Realize automated picking and inspection of bulk parts by using AI without temporary placing




The transfer of works between manufacturing processes, such as loading of works onto machine tools and hauling them between different tasks still rely heavily on human labor and is a field with a strong demand for automation. By automating the task of picking up machine parts that are required between these manufacturing processes,  the efficiency of workflows between processing tasks and of the whole assembly line are improved.

In this client’s case, while they already had an automated picking system, temporary placing was still required before the works could be loaded onto the jigs, which resulted in long lead times. Additionally, they were conducting visual inspections for the sealer application on the works that were carried over, which resulted in issues with inspection cost and quality.

Issues associated with loading of works onto machine tools and moving them between processes

The following 3 issues were identified.

  • The temporary placing was required for the picking system, and long lead times occurred until the works were set on the target jigs.
  • As the sealer application inspection is done visually, quality control was unreliable and had costs associated with employing inspectors.
  • With the current picking system, it is necessary to make complicated settings every time the workpiece to be picked is switched.

Realize accurate picking and inspection automation by our algorithm

Based on data from 3D cameras, our company’s picking algorithm detects the orientation and angle of the parts, enabling the automatic picking system to adjust itself to various different types of work. Additionally, using an algorithm that converts the 3D camera’s coordinate system to the robot’s, it can accurately estimate the position of the parts relative to the robot arm, enabling it to carry parts to the desired setting position on the jigs without temporarily placing.

To automate the inspection process, we used an algorithm (Semantic Segmentation) that correlated data from every pixel of the photo taken by the camera to scan the application state. The results of the scan are used to evaluate if they satisfy each item on the checklist (thickness, length, position), effectively automating and quantifying the inspection

Eliminating need for temporarily placing and visual inspections reduced workload and manpower while improving line productivity

We created a picking system that enables robots to accurately pick up works bulk parts. Eliminating the need for temporary placing which was required previously also made it possible for the robots to directly set the works onto the jigs without swapping, which helped reduce workload. In addition, the automation of inspections after the work is carried over reduced the required manpower and made the quality control inspections quantifiable.