Abstract
In this paper, classification of ball bearing faults using vibration signals is presented. A review of condition monitoring using vibration signals with various intelligent systems is first presented. A hybrid intelligent model, FMM-RF, consisting of the Fuzzy Min-Max (FMM) neural network and the Random Forest (RF) model, is proposed. A benchmark problem is tested to evaluate the practicality of the FMM-RF model. The proposed model is then applied to a real-world dataset. In both cases, power spectrum and sample entropy features are used for classification. Results from both experiments show good accuracy achieved by the proposed FMM-RF model. In addition, a set of explanatory rules in the form of a decision tree is extracted to justify the predictions. The outcomes indicate the usefulness of FMM-RF in performing classification of ball bearing faults.
Highlights
Condition monitoring of machinery is becoming increasingly important in modern maintenance
A multi-stage algorithm based on artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) model is used for classification, with results showing the proposed method is effective [23]
While the results of Fuzzy Min-Max (FMM)-Random Forest (RF) acquired the highest accuracy rate with the smallest standard deviation, Classification and Regression Tree (CART) [35] on the other hand had the simplest network with five leaf nodes
Summary
Condition monitoring of machinery is becoming increasingly important in modern maintenance. One of the most commonly used predictive maintenance technology is vibration monitoring, due to the amount of machine conditions information that is provided [2]. Condition monitoring in the rotating machines of the industry uses accelerometers and vibration transmitters in order to acquire data [3,4,5]. Pattern recognition is the central task in the machine condition monitoring, with various solutions reported [6,7,8,9] It first looks at information from multitude of sources, such as transducer signals. The contributions of this paper are two-fold: the use of a hybrid intelligent model for detection and classification of real-world roller ball bearing faults as well as detailed investigations in the use of a set of power spectrum and sample entropy-based features for performing this task.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
