Qualodoromat -

AI methods for odor evaluation of plastics

The ZIM joint project "Qualodoromat" (originated from the ZIM network KI-MAP) researches methods of automated odor characterization of plastic samples using artificial intelligence and robotics.

ZIM-funded joint project Qualodoromat

  • Problem

    Nowadays, testing of plastics and polymeres, especially with regard to odor, is carried out manually by experts. This requires a great deal of experience and time. In addition to the subjective evaluation of the odor impressions, such odor tests sometimes represent a high level of stress for the human nose. Furthermore, the health aspect should not be neglected, both with regard to the ingredients and the current state of health of the expert team. This leads to limited test capacities, since human-sensory odor tests cannot be realized "on an assembly line".

  • The Partners

    This is where the research consortium consisting of the company Genie Enterprise, the non-profit KIMW-Forschungs-GmbH and the Umwelt Campus of Trier University of Applied Sciences (Departments of Environmental Planning/Environmental Engineering and Environmentally Sound Production Processes & Industrial Robotics) comes in and focuses on a technical solution: the "artificial nose" for plastic samples.

  • Research

    Based on sensor measurements of plastic samples, which are annotated by experts of the non-profit KIMW-Forschungs-GmbH, a modeling of odor profiles is performed by us. For this purpose, various current AI methods are used and adapted. This should make it possible to classify plastic samples automatically into odor profiles and to derive objective odor grades from them. This is done in accordance with currently valid norms and standards (e.g. PV 3900 and VDA 270). The quality of the results should be comparable to those obtained with the help of a human nose by a specially trained person.

HOW

 

Trier University of Applied Sciences is responsible for the development of a research demonstrator, which realizes both the sample handling and the sensory measurement of the plastics. For the odor measurements, a multi-sensor device is being developed that is adapted to the specific requirements of the plastics. Furthermore, robotic components will be adapted for a technical preparation and feeding of the samples. The combination of sensor system and robotics allows an automated handling of the samples with reduced manual effort. This makes it possible to measure plastic samples in larger quantities and with a consistent quality. At the same time, Trier University of Applied Sciences, in cooperation with the non-profit KIMW-Forschungs-GmbH and Genie Enterprise, will build up a database for recording the reference data for the plastic samples, which will be available for training machine learning methods.

Furthermore, Trier University and Genie Enterprise will address aspects of active sensor control. Here, methods of Adaptive and Behavioural Learning will be applied to the adaptation of sensor parameters.

By combining AI methods, sensor adaptations, supporting robotics and expert knowledge from odor testing, the ZIM joint project "Qualodoromat" can create an objective and high-performance system for odor evaluation of plastics.