Our client needed a solution to check the assembly of plate heat exchangers, which are complex devices consisting of many visually similar but functionally different plates. Faulty parts in a heat exchanger can lead to a malfunction in the final product – the car, and result in costly recall claims for our client.
Assembly errors are very easy to make and extremely difficult to detect, due to a number of reasons. The identification elements of each part are too small, making it impossible for a human to recognize them. There are also a great variety of components that can be missed or mistaken for each other. The task is challenging even for a camera, as the aluminium pieces are mirror-like and their light reflection hinders optical inspection.
We created AI-enabled vision to accurately identify each heat exchanger component and check the correctness of the entire assembly process. In addition, the neural network training that could previously only be achieved by programming experts is now accessible to anyone, so components or processing methods can be added or modified as the production needs dictate.