Fault Diagnosis of Reciprocating Compressors Based on Thermal Imaging and Support Vector Machines

Reciprocating compressor (RC) is widely used in industry because of its wide application pressure range, high thermal efficiency and strong adaptability. In order to avoid the possible casualties or property losses caused by the fault of a RC, it is of great significance to carry out efficient fault diagnosis for RC. The common fault types of RC usually lead to the variation of pressure range and the amount of air intake, so different faults will lead to the changes in the distribution of the whole temperature field, which provides the possibility to distinguish the fault types of RC by using infrared thermal imaging. In this paper, three kinds of faults were simulated in an uncontrolled temperature environment. The temperature distribution of a RC was captured by a infrared camera remotely. During the experiment, the average grayscale values of six main components were calculated in real time to form a 6-dimension vector which represent the temperature distribution. Support Vector Machines (SVM) was then used for classifying the differences of the temperature distributions, and the results demonstrated that variation of the average temperatures of six main components aided by SVM has great potential for fault diagnosis of RCs under various operating conditions and ambient temperature.