In a highly automated manufacturing plant, the reliability of manufacturing equipment is critical for normal operation. A sudden machine breakdown can bring unexpected downtime, shorter lifespan of equipment, and lower operational efficiency. Breakdowns can lead to defective parts and consume extra energy—issues that are undesirable from an environmental standpoint—and also erode productivity and increase costs. To improve machine tool reliability, a machine may be continuously monitored to track its health condition. Monitoring a machine often provides large amounts of data that must be processed to distill useful information. Electric motors are found in many pieces of common manufacturing equipment. Deep learning methods can be combined with data collected on motors, e.g., acceleration time-frequency data, to identify motor condition. In this paper, three state-of-the-art deep learning architectures are evaluated for their ability to effectively monitor motor condition. Experiments are performed on a lab-scale motor test bed to secure condition data for several common motor faults. Tri-axial acceleration data are collected and converted into 2D images (spectrograms) using the power spectral density function. Some of these experiments are used to tune the deep learning algorithms, and others are used to test the proposed monitoring methods. The relative performances of the architectures are assessed, and it is demonstrated that the use of time-frequency images within a deep learning context can efficiently handle large amounts of data and effectively monitor the motor condition.
Deep learning Motor condition monitoring Convolutional neural network Motor test bed
This is a preview of subscription content, log in to check access.
This work is supported by the Wabash Heartland Innovation Network (WHIN).
Compliance with Ethical Standards
Any opinions, findings, conclusions, and/or recommendations expressed are those of the authors and do not necessarily reflect the view of the WHIN.
Le Cun Y, Boser B, Denker JS et al (1990) Handwritten digit recognition with a back-propagation network. In: Advances in Neural Information Processing Systems 2. Morgan Kaufmann, pp 396–404Google Scholar
Eren L, Turker I, Kiranyaz S (2019) A generic intelligent bearing fault diagnosis system using compact adaptive 1D CNN classifier. J Signal Process Syst 91:179–189CrossRefGoogle Scholar
Mahmood F, Toots M, Öfverstedt L-G, Skoglund U (2018) Algorithm and architecture optimization for 2D discrete Fourier transforms with simultaneous edge artifact removal. In: International Journal of Reconfigurable Computing, pp 1–17CrossRefGoogle Scholar
Sharma N, Jain V, Mishra A (2018) Analysis of convolutional neural networks for document image classification. In: Procedia Computer Science. Elsevier B.V., pp 377–384Google Scholar
Scherer D, Andreas M, Behnke S (2010) Evaluation of pooling operations in convolutional architectures for object recognition. In: Artificial Neural Networks - ICANN 2010 - 20th International Conference, pp 92–101CrossRefGoogle Scholar