Around 60,000 compressor units are in operation at German companies. Together, they use up 16.6 terawatt hours of power per year and generate related costs. However, the majority of these units have leaks in their connecting parts or tiny holes and kinks in their pipes, which significantly impairs their efficiency. The Fraunhofer Institute for Manufacturing Engineering and Automation IPA in Stuttgart has calculated that if leaking pipes and components were replaced, the costs for these units could be reduced by 30%.
The problem is detecting leaks in the first place. The IPA proposes using AI to perform this task , and to this end has built a demo prototype as a first step. The demo system measures the degree of pressure with which the air passes through the pipes, identifies the flow rate, the position of the actuators, and the condition of the valves, and captures the ultrasound signal. This data is saved in the cloud and creates the basis for training the AI, which will then not only be able to identify and localize any leakages, but at the same time display a description and the serial number of the leaking component.