Thermoelectric Monitoring Main Bearing Using TrAdaboost Learning Algorithm in Diesel Engine

In response to the difficulty of monitoring the main bearing wear of the same type of diesel engines with different operating conditions by the thermoelectric method, the boundary conditions of bearing pads were determined using a variable load test on a sapphire sliding bearing test bench; the amplitude of the thermoelectric signal appeared at the maximum bearing load. The 15-h wear test revealed that the thermoelectric signal amplitude increased with an increase in the wear time. The wear simulation test of the main bearing was carried out on a diesel engine bench; it verified the corresponding relationship between the thermoelectric signal and the main bearing load. Upon an increase in the wear of the main bearing, the thermoelectric signal showed an upward trend, and the thermoelectric characteristic value of the corresponding adjacent cylinder of the bearing was extracted. Based on the TrAdaBoost learning algorithm, a diagnosis model for the diesel engine main bearing wear was constructed. A verification of the simulated main bearing wear test samples revealed that the diagnosis accuracy was more than 95%. The results showed that the diagnosis model could realize the diagnosis knowledge migration among the different load conditions of the diesel engine.