Process Mining in Battery Life Cycle

(a) Situation faced: A charged battery for the IoT-enabled bike locks is essential for the bike rental process in the DB Connect bike-sharing system. To ensure the availability of the bikes, the battery must be charged at regular intervals in addition to the classic maintenance processes. If the battery dies, a bike is no longer rentable to the customer. As the rental stations are located outside, the bikes are exposed to external influences such as temperature, humidity, and sunlight, all of which directly impact the charging cycle of the batteries. The project presented provides an initial idea of how we can support decisions in future maintenance processes to achieve an optimal cost-benefit ratio. (b) Action taken: Bike maintenance is a process that can be well represented in discrete, finite events. To analyze the performance of the process, we focused on the voltage curve of the battery data, although the course of this is difficult to predict. Influencing factors here include weather conditions, frequency of use, and the age of the battery. Typical charging cycles were modeled from the data, forming the basis on which we optimized the maintenance process. (c) Results achieved: The analysis gives us transparency on the number and duration of maintenance cycles per bike in terms of battery charge, including external factors including frequency of use and weather-related variations. In addition, knowing the threshold of the critical voltage allows us to predict and prevent bike failure. This was the starting point for an overall optimization of the predictive maintenance processes. (d) Lessons learned: When presenting our project, we primarily focus on the psychological aspects. By using process mining, we were able to discuss processes transparently with the company process owners and did not have to limit ourselves to the subjective perceptions of the experts, as was often confirmed by the visualizations.