Multiple classification of the acoustic emission signals extracted during burn and chatter anomalies using genetic programming

Abstract

This paper looks at the multiple characteristics and investigations of two grinding anomalies: grinding burn and grinding chatter. Most researchers just look at one phenomenon anomaly investigation and classification as a single entity. This, however, is not flexible for efficient monitoring of grinding manufacture where many unwanted anomalies are required to be monitored. This paper provides two experimental set-ups to obtain signatures for both chatter and burn phenomena. This is significant to any manufacturing environment due to both anomalies providing surfaces that cannot be accepted when scrutinised for quality purposes and, therefore, play an integral part into the manufacturing process. This paper also looks at the novel evolution classification techniques and specifically Genetic Programming (GP) which provides the multi-classification paradigm. GP is based on evolution: ‘Survival of the fittest’ strategy and surfers from the ‘curse of dimensionality’ where techniques such as statistical feature n-dimensional reduction and Independent Component Analysis provides the blind separation and compression of salient point’s reference to specific signal phenomenon. The demarcation between each of the phenomenon was identified from acoustic emission signals being converted to the frequency–time domains using Short Time Fourier Transforms (STFT). Other digital signal processing techniques were used and discussed; however, the more up-to-date and successful tests only required STFT.