Datasets for condition monitoring

Manufacturing companies continuously try to increase their productivity, among others by avoiding machine down times. The latter involves considerable costs because of the resulting loss of turnover. Monitoring the condition of, for instance, bearings and gears, plays a vital role in the maintenance program of rotating machines. Early fault detection could allow to move from a time based preventive maintenance program to a condition based predictive maintenance strategy and reduce unexpected machine downtime and cost. Moreover, it can avoid the unnecessary and costly replacement of healthy machine components parts.

A key barrier in the wider adoption of condition monitoring is the lack of large and reliable data sets about the full lifetime of bearings and gears in machines as basis for algorithm development, testing and validation. This is, firstly, due to the fact that failing machines are typically very rare. Secondly, because acquisition of relevant data about bearing faults and lifetime requires a large effort and takes a long time. Only very few public data sets are available, but they are recorded under limited operating conditions. Large and well documented data sets could be used for the development, testing and validation of bearing fault diagnostic and prognostic methods or in benchmark studies to compare methods.