Pumps, as core pieces of equipment in ships, are installed in the engine room to supply refined oil to the engine. Pump failure causes critical problems for ship operations. Therefore, failure-monitoring-based diagnosis technology is an essential requirement in the shipbuilding industry. For this purpose, a database containing information about the failure states depending on the main cause of the failure cases of the pump needs to be developed. In the present study, failure causes of pumps based on actual accident records were quantitatively analyzed. Then, failure modes for the bearing, coupling, sealing, and screw, which are the core parts of the oil pump, were determined. Test infrastructures for the oil pump were developed to obtain normal and abnormal data considering diverse operating conditions. Based on the vibration data from the accelerometer installed on the test infrastructures, the frequency of failure was analyzed through Fast Fourier Transform (FFT). In addition, more precise results were obtained by performing Short-Time Fourier Transform (STFT) for the FFT results that indicated severe failure. Finally, over 200 data entries were accumulated on the core parts of the oil pump, considering normal as well as abnormal operating conditions. The database constructed in this study is expected to help in investigating failure diagnosis and prediction of algorithm models for ship management.