Assessment of carbon reduction through AI methods in inspection after reverse logistics

Although it is a goal to reduce the workforce and energy consumption required to digitize and learn Artificial Intelligence (AI) methods, it is vital to assess the balance of energy consumption versus the efficiency benefits through AI methods for reverse logistics sorting processes. Therefore, the aim of the paper is to describe how to find the return of investment in energy used for digitization and artificial intelligence. At first, energy consumption drivers introduced to the sorting process for digitization are assessed. In addition, the energy consumption of the training of AI methods and other sophisticated statistical models was monitored. Afterward, the authors discuss and select which metrics suit the comparison between energy costs of algorithms and show the influence of design decisions in artificial intelligence. An analysis shows that it is possible to achieve high recognition performance but keep energy consumption in mind.
Finally, it is shown that in an inspection use case, we can achieve a fast break-even of energy use consumption for digitization and AI used for improved processes is achievable. The demonstration compares the energy costs of more properly sorted remanufactured products compared to the opportunity costs of energy consumption for a new production of respective products.