This article addresses an improvement of a classification procedure on cracked rotors through Deep learning based on convolutional neural networks (CNNs). At first, a cracked rotor-bearing system is modeled by the finite element method (FEM), then throughout its start-up, the related time-domain responses are calculated numerically. In the following, as a pre-processing stage, continuous wavelet transform (CWT) and Short-time Fourier transform (STFT) are applied on the three various health conditions, i.e. without crack, shallow-cracked, and relatively deep-cracked shafts. The plots of CWT’s coefficients and STFT’s in these various classes are used as the input dataset in Deep learning based on CNNs and the three classes are introduced as the output. AlexNet with 25 layers is employed as the network. The results of the testing phase demonstrated that not only this expanded method has a reasonable capacity in the classification of cracked and healthy rotors, but it also can classify cracked rotors with different crack depths with a negligible error.

The positive benefits of early faults detection in rotating systems have led scientists to develop automated methods. Although unbalancing is the most prevalent defect in rotor systems, this fault normally is accompanied by other defects such as crack. In this article, an effective self-acting procedure is addressed in identifying shallow cracks in rotor systems throughout the steady-state operation. To classify rotor systems suffering cracks with three various depths, firstly, healthy and cracked systems are modeled by employing the finite element method (FEM). In the following, systems' vibration signals are calculated in different situations numerically; for pre-processing stage, the persistence spectrum is implemented. Finally, by using a supervised convolutional neural network (CNN), rotor systems are classified by regarding the crack depths. The result of the testing step revealed that this hybrid method has rational capacity in distinguishing shallow cracks in steady-state operation where many other methods are somehow powerless.

Parallel with significant growth in industry, especially mysteries related to energy engineering, condition monitoring of rotating systems have been experiencing a noticeable increase. One of the prevalent faults in these systems is fatigue crack, so finding reliable procedures in identification of cracks in rotating shafts has become a pressing problem among engineers during recent decades. While a vast majority of cracked rotors can operate for a specific period of time, to prevent catastrophic failures, crack detection and measuring its characteristics (i.e. size and its location) seem to be essential. In the present essay, a hybrid procedure, consisting of Deep Learning and Discrete Wavelet transform (DWT), is applied in detection of a breathing transverse crack and its depth in a rotor-bearing-disk system. DWT with Daubechies 32(db32) as wavelet mother function is applied in signal noise reduction until level 6, also its Relative Wavelet Energy (RWE) and Wavelet entropy (WE) are extracted. A characteristic vector that is a combination of RWE and WE is considered as input to a multi-layer Artificial Neural Network (ANN). In this supervised learning classifier, a multi-layer Perceptron neural network is used; in addition, Rectified Linear Unit (ReLU) function is exerted as activation function in both hidden and output layers. By comparing the results, it can be seen that the applied procedure has strong capacity in identification of crack and its size in the rotor system.