Transitional data for estimation of gearbox remaining useful life

Past failure tests conducted on mechanical equipment were directed towards the goal of determining the number of cycles to some measure of failure and/or a statistical measure of overall reliability such as a mean time to failure. True Condition-Based Maintenance (CBM), however, requires the identification and tracking of the sensor observables capable of detecting the faults and the ability to relate these variables to the overall health and remaining useful life of the machine. Thus, as part of a larger-scale research program in CBM for machinery, a Mechanical Diagnostics Test Bed (MDTB) was constructed to provide data on a commercial transmission as its health progresses from new to faulted and finally to failure. The paper presents a full description of the MDTB research station and its instrumentation. The generation of continuous run/good-to-bad gearbox transitional data will be used to determine appropriate data fusion and approximate reasoning techniques that result in the identification and detection of precursors to failure. The MDTB will also provide a test and evaluation vehicle for emerging prognostic time-to-failure/remaining life prediction algorithms and advanced sensors developed either at Penn State or by other participating researchers.