An innovative system for automated spare parts identification, developed by ITK Engineering and Bosch Cognitive Services, reduces production downtime and increases efficiency. By using cloud technologies and AI based on IoT data, spare parts can be identified quickly and accurately. This solution not only saves time, but also opens up new business opportunities.
The challenge: Reducing complexity and downtimes
Production downtimes are costly. If a production line comes to a standstill due to a defective component, it must be removed and the corresponding spare part quickly and easily identified and procured. Effective spare parts management is therefore crucial for minimizing downtimes and ensuring productivity.
Production lines often contain tens of thousands of different types of parts. Conventional methods for part identification are therefore time-consuming and often heavily dependent on the experience of employees. Incorrect or delayed identification can lead to significant production downtime, which is costly and impairs productivity. Companies need a comprehensive and well thought-out strategy to successfully master these challenges.
Holistic data-driven development can be a solution: By not only generating data, but also gaining intelligent insights from it using machine learning, companies create direct added value. The development service provider ITK Engineering offers exactly the right range of services for these requirements.
The solution: Automated spare parts identification and scalable data pipelines
A central element of ITK Engineering’s solution is scalable data pipelines that enable the effective processing of large data streams. The focus here is on flexible data infrastructures that can be operated both locally and in the cloud in order to meet the specific requirements of the respective application.
A cloud-based AI solution is used for automated spare parts recognition. At its core is a neural network that enables spare parts to be identified quickly and precisely using photos.
In order to capture and display the spare parts easily, a system was developed that automatically photographs spare parts from different perspectives and feeds the images into a database. These images then serve as the basis for training the AI models.
The solution is trained with uploaded photos in the cloud and executed locally with embedded AI on an industrial computer. Maintenance staff can take a photo of a defective part with a smartphone or tablet or upload an existing image to the cloud. Detection also works with different backgrounds, low image quality and dirty or damaged components. Within seconds, the system provides a list of possible spare parts with ID and storage location.
The result: Increased efficiency and new business opportunities
Automated spare parts identification enables spare parts to be identified quickly and reliably, drastically reducing downtimes and the associated costs. The solution is able to differentiate between a variety of up to 20,000 classes of spare parts.
In future, the high degree of automation will enable new use cases to be trained with a single mouse click – even by employees without machine learning or cloud expertise. If spare parts recognition is made available to other companies as a service, they also benefit and the reach of the solution increases.
