26 / 2015-08-11 15:45:31

Fault Diagnosis of Marine Reefer Containers Based on PCA-SVM

5604,6394,3097,7527

There are numerous variables in the fault samples of marine reefer containers and collinearity exists among them, so directly use them to train the fault diagnosis model will lead to bad performance. In the paper, the Principal Components Analysis method was used to extract the main information of the faults data and the pretreated data was then used as the model input. SVM method was chosen to build the multi faults diagnosis model, and a fault diagnosis model of reefer containers based on “One versus Rest” SVM was set up. The experimental results showed that the multi faults diagnosis system of reefer containers based on PCA-SVM had high fault classification accuracy of more than 98.4% and fast diagnosis speed. Comparing with the SVM model without PCA preprocessing, fault diagnosis accuracy increased by more than 1.61% and model training time fell nearly 10 ~ 30 times.