Fault Detection in Centrifugal Pumps using Fuzzy Logic and Artificial Immune Network

Failures and sudden breakdowns of centrifugal pumps, which are one of the most important and most widely used rotary machinery in various industries, can cause a lot of problems. Thus, precise and on time condition identification and fault detection of these machinery are of great importance. In order to increase the accuracy and speed of condition identification, different machinery and intelligent methods such as neural networks, fuzzy logic or a hybrid of them are used. In this study, an artificial immune network that is inspired from the human immune system is combined with the ANFIS method for pattern reorganization and fault detection in centrifugal pumps. For this purpose, first the experimental data for different states of the system are collected by creating a test setup and performing tests. After that, different statistical features are extracted from vibration and current signals in time, frequency and time-frequency domains; also, three more important features were selected by IDE and PCA methods. Next, data were classified by Kmeans, SVM and a proposed hybrid method and their errors were investigated. The results showed that the new method has high accuracy in fault detection and is even able to detect new conditions correctly without further training