Artificial intelligence for sewer condition assessment
The public sewage system in Germany has a total length of approx. 580,000 km. Several billion euros are invested in its maintenance every year, with the assessed sewer condition being the main
basis for decision. In order to assess the condition of the sewer, it is flushed, driven through with a video camera, and any damage is identified in a manual process. However, evaluating the sewer inspection images is a very tedious, subjective and error-prone undertaking. On average, 25 % of the damage is either not detected at all or is incorrectly coded (Dirksen et al., 2013). This high level of subjectivity leads to a high degree of uncertainty and non-reproducible results, which limits the application of data-based decision models to optimize sewer maintenance.
The software developed by ETH Zurich spin-off company Hades uses artificial intelligence to automatically detect any damage in sewer inspection images. Instead of going through the videos and identifying damage manually, all that is required is to upload the videos to the web-based app, which then detects and codes the damage. This process is not only faster but also permits objective, less error-prone and consistent evaluation. The data quality of the condition record is optimized as a result, allowing the implementation of models for data-based sewer maintenance.