We are working on a predictive maintenance system in the area of assembly line maintenance incl. hardware components for a well-known German logistics company. Audio data and the positions of the sorting system's individual cells are recorded with sensors specially developed for this purpose and then analysed for maintenance-relevant anomalies in high resolution using statistical methods. Unplanned breakdowns can thus be prevented and efficiency increased. Above all, however, the use of automated maintenance prognosis considerably reduces personnel requirements as only those few elements of the sorting system specified by the prognosis algorithm have to be checked, instead of all of them.
The aim of the project was to develop a predictive maintenance system based on the evaluation of audio data using data science methods. The software should recognise the degree of wear of the cells on the conveyor belt based on the recorded vibrations and generate a maintenance case. Particularly challenging were the handling of noise disturbance and the significantly different mechanical design of many of the sorters.
With the help of proven data science methods, a predictive maintenance system (hardware and software) was developed in the area of conveyor belt logistics. The system recognises maintenance-relevant abnormalities in the wagons and generates an automated notification in the management interface. This significantly reduces manual inspection efforts, ensures good planning of the maintenance process with graduated advance warnings and prevents avoidable failures.
In the future, the maintenance forecasting system will be rolled out worldwide at the logistics service provider's various locations. The focus will be on adapting the forecasting algorithm to other sorter types and remotely configuring/updating the sensors.
Interested? Then just take a look at our prototype.
Software and hardware development, setting up the operating environment
03/2022 – today
400 person days