Transfer Learning-Based Fault Diagnosis Method for Marine Turbochargers

To address the issues of high cost marine turbocharger fault simulation testing and difficulties in obtaining sample data, a multi-body dynamics model was developed. The approach used to acquire vibration signals. result shows that amplitude 1× signal power spectrum drops as bearing surface roughness increases. However, 2× 9× spectra increases TrAdaBoost transfer learning method is develop diagnosis model. validation results 2040 simulated samples reveal when desired number 20, diagnostic has an accuracy 87%. When 40, model’s 96%. may perform information between actual