Improve automotive inspection accuracy and cut down assembly inspection time with LandingLens
Efficiency and accuracy are crucial factors in the automotive manufacturing industry, and this globally renowned automotive company (manufacturer) which stands at the forefront of the industry delivering unrivaled quality and innovation, recognized the need for automation in their manual inspection of their vehicle doors. To address this challenge, they sought the expertise of FPT, a global IT consulting and service provider and a Landing AI Platinum Partner. FPT and Landing AI collaborated to implement a game-changing computer vision solution that improved accuracy from 50% to 99.7% and cut down inspection time from 3 minutes to 3 seconds.
In this article, we explore how LandingLens, the cloud-based computer vision platform from Landing AI, can optimize your manufacturing process, saving you time and money.
Challenges with Manual Inspections
The manufacturer was facing a few challenges with inspecting screws inside vehicle doors. Screws could potentially be installed in the wrong position, cross-threaded, or missing altogether. Additionally, these screws were installed inside the vehicle door, so operators viewed X-ray scans of each door.
Despite human inspectors dedicating up to 3 minutes per door inspection, manual inspections only yielded 50% to 60% accuracy. They recognized the need for a more efficient and reliable solution.
The LandingLens Solution
FPT, in collaboration with Landing AI, implemented a cutting-edge computer vision solution to the production line. By utilizing the LandingLens platform, FPT uploaded X-ray images of both correctly and incorrectly installed screws. The team then labeled the automotive images, distinguishing them as either “Good” or “Not Good” using a Classification project in LandingLens. The computer vision model was trained on this labeled dataset, fine-tuned as necessary, and finally deployed.
The implementation of LandingLens delivered exceptional outcomes for the manufacturer. The system now provides operators with a clear “OK” or “NG” (“Not Good”) output, achieving an impressive accuracy rate of 99.7% within a rapid inspection time of 2-3 seconds per door.
By automating the inspection process, they significantly reduced the number of escaped defective door parts. This proactive approach eliminated the need for costly quarantine measures and reactive reinspection, resulting in annual savings exceeding $200,000. Moreover, the LandingLens solution effectively mitigated door recalls and potential safety risks, preserving the reputation of the manufacturer’s brand.
Inspectors Now Clearly See if the Door Passes (OK) or Fails (NG for “Not Good”) Inspection
FPT and Landing AI: A Winning Combination
The manufacturer’s decision to partner with FPT and Landing AI was motivated by their dedication to understanding and fulfilling their specific requirements. The collaborative nature of FPT and Landing AI, coupled with their ability to incorporate feedback from the manufacturer, played a crucial role in the success of the project.
The ongoing support and exceptional service provided by FPT and Landing AI set them apart as trusted partners. The manufacturer expresses their intention to continue working with FPT and Landing AI for future computer vision inspection projects and use cases.
Fast Deployment, High Accuracy, and Future Expansion
One of the standout advantages of the LandingLens solution was its swift deployment, requiring just one month for implementation. This quick turnaround allowed the manufacturer to witness immediate benefits and optimize their operational efficiency.
With a significant improvement on inspection accuracy of 99.7% and substantial cost savings achieved, the manufacturer plans to roll out the LandingLens solution across North America facilities. This expansion demonstrates the manufacturer’s confidence in the capabilities of Landing AI’s computer vision platform.