Energy harvesting based predictive maintenance system for centrifugal pumps in swimming pools

Abstract:
An unexpected machine failure often brings entire processes to a standstill and causes high costs. To counteract this, machines are often maintained ahead of time, which leads to high costs. Predictive maintenance offers a good solution here, as it detects emerging defects at an early stage, and maintenance can then be carried out as needed. This requires wireless sensor nodes, which must be supplied with energy. Energy harvesting is a good solution here, as there is no need to change batteries or complicated cabling. We present an energy harvesting based predictive maintenance system, which was developed in a joint research project for the centrifugal pumps of the company Herborner Pumpen. By cleverly integrating thermocouples between the hot side of the motor and the cooling circuit of the system, power in the range 1 mW can be generated, which is sufficient to operate a sensor system. In addition to monitoring the integrated pump filter, it is also possible to monitor other critical pump components. The system includes a power management system which can be flexibly adapted to the requirements of the application. A variety of sensors as well as different types of energy harvesters can be used. The system stores surplus energy in a battery to ensure pump monitoring even during periods without ambient energy supply. Sensor data is transmitted wirelessly to a cloud where it is analysed using artificial intelligence. The time series analysis based on the sktime framework enables a runtime-efficient analysis of the filter basket.