Mechanical faults in rotating machinery dataset (normal, unbalance, misalignment, looseness)
Description
If you use this dataset, please cite: Lucas Costa Brito, Gian Antonio Susto, Jorge Nei Brito, Marcus Antonio Viana Duarte. Fault Diagnosis using eXplainable AI: A transfer learning-based approach for rotating machinery exploiting augmented synthetic data. Expert Systems with Applications 232(4):120860. DOI: 10.1016/j.eswa.2023.120860 The dataset was developed aiming to address classic faults in rotating machines such as: unbalance, misalignment and mechanical looseness, in addition to the normal operating condition. The faults were introduced on a test bench, being: motor, frequency inverter, bearing house, two bearings, two pulleys, belt and rotor (disc). 20 tests were performed, 5 for each condition (normal, unbalance, misaligment and looseness). Each test consists of 4 sets of 420 signals collected continuously, each file consisting of 25,000 points with the sampling rate set at 25 kHz (420 signals per accelerometer). Resulting in the end of all tests, in a total of 8400 signals per accelerometer. The sequence of tests was randomly defined. Before starting any test, the bench was dismantled and returned to normal operating condition, to later introduce the fault. The experimental procedure allows variations to occur, making the tests closer to industrial reality. Each file has (4, 25000), where 4 represents the position of the accelerometers, and 25000 the number of points. The accelerometers were positioned in such a way that: 1 - Vertical side of the coupling (pulley) 2 - Horizontal side of the coupling (pulley) 3 - Vertical side opposite the coupling (disk) 4 - Horizontal side opposite the coupling (disk) The sequence of tests and operating conditions are: - Test 01: Normal Condition - Test 02: Misalignment - Test 03: Normal Condition - Test 04: Unbalance - Test 05: Mechanical Looseness - Test 06: Normal Condition - Test 07: Unbalance - Test 08: Mechanical Looseness - Test 09: Normal Condition - Test 10: Misalignment - Test 11: Misalignment - Test 12: Mechanical Looseness - Test 13: Misalignment - Test 14: Mechanical Looseness - Test 15: Unbalance - Test 16: Misalignment - Test 17: Normal Condition - Test 18: Unbalance - Test 19: Mechanical Looseness - Test 20: Unbalance
Files
Steps to reproduce
Bench description: - Motor: Three-phase induction motor, B56 B4, Manufacturer Eberle, nominal speed 1650 rpm, power 0.09 kW, voltage 220 V, current 0.70 A, bearings 6200 ZZ. - Frequency Inverter: Vector Inverter, model CFW300, manufacturer WEG. - Driven Pulley (Disc Side): model A80, external diameter 80 mm, internal diameter 53 mm. - Motor Pulley (Motor Side): model A60, external diameter 60 mm, internal diameter 37 mm. - Belt: model V - A23, top width 12.7 mm, outer perimeter 630 mm, inner perimeter 580 mm. - Rotor (Disc): diameter 149 mm, 36 holes, thickness 6.5 mm. - Bearings 1205, bearing house SN505, bushing H305. For the acquisition of the signals were used: 04 accelerometers PCB 352C33 (2 in each bearing), mounted in the vertical and horizontal positions (y- and x-axes), 01 Power Supply PCB 482A20, 01 acquisition board Hi-Speed USB Carrier, NI USB-9162. A Python language program was developed to collect the data. All accelerometer sensitivities were adjusted according to the calibration chart. The rotation was kept constant with a value measured on the shaft of approximately 1238 rpm. In addition to studying the dynamic behavior of faults in rotating machinery, the dataset can be used for training artificial intelligence models for: fault diagnosis, anomaly detection, transfer learning, etc.