Pneumatic Cylinder Diagnostics using Classification Methods

Air leakage detection in a simple pneumatic system utilizing artificial neural networks and support vector machine classifiers was investigated. Training and test data for the built classifiers was experimentally collected introducing artificial leakages. Single and multi level classifier structures were implemented and their performance compared by means of classification error rate. Based on the available experimental data set it was found out that multi level classifier based on support vector machine subclassifier outperformed other classifiers by exhibiting the lowest 2% classification error rate.