Recurrent neural network based vibration data classification

Mohd Zamberi, Nurul Nadheerah (2020) Recurrent neural network based vibration data classification. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Recurrent neural network based vibration data classification.pdf - Submitted Version Download (2MB)

Abstract

Vibration is the main part in the machinery, the type of vibration is the important for the best output result and to avoid major breakdown for the production. Vibration data classification has been research mainly focusing of improving the efficiency and the operation of the production. The dataset are mixed with numbers of vibration data with different type of vibration produced from the cutting machine, which bring difficulty to analysis the dataset for accuracy result. Other than that, the sampling time of analysing the vibration data also affect the accuracy result. The sampling time that will be analysed is between 10 microseconds to 50 milliseconds. Recurrent neural network (RNN) is a type of artificial neural network which can be embedded with multiple of time sequence data. The capacity of RNN had been prove outstanding for entering time relevance about the time sequence data. This paper proposed a method for recurrent neural network based vibration data classification. The RNN algorithm will be build using the MATLAB software. From the total datasets it will be divided into three parts for training, validation and testing. The final result of this analysis will be determine by the confusion matrix which shows the accuracy result of the data classification. The final accuracy result of 10 microseconds is 75.7% and for 50 milliseconds is 65.4%.

Actions (login required)

View Item