In the actual working process, the source of vibration signal is not only the rotor itself, so the detected vibration signal will become complicated. This complex signal makes it difficult to accurately measure the existence of crack. In this paper, a novel method, which includes complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and continuous wavelet transform (CWT), is proposed to analyze the cracked rotor-rolling bearing system. The CEEMDAN-CWT successfully separates the vibration signal of the rotor itself from the original signal and provides results similar to the simulation signal. At the speed below 2000 rpm, the 2X frequency difference between cracked rotor and healthy rotor in CEEMDAN-CWT spectrum is about 1, while the difference of FFT spectrum of original signal is about 0.6, which shows the superiority of the novel method in extracting rotor vibration signals from complex vibration signals.

The modernization of electrical power systems is reflected through the integration of renewable energy resources, with the ultimate aim of creating a carbon–neutral world. However, this goal has brought new and complex challenges for the power system, with one of the most crucial issues which is the reduction of system inertia. The decrease in system inertia has led to severe difficulties in maintaining frequency stability. As a result, power system operators must continuously monitor the system inertia and when necessary to activate appropriate preventive measures, ensuring a reliable and secure operation of the power system. Fortunately, wide-area monitoring systems can provide the necessary measurements to monitor and analyze system behavior, assisting system operators in undertaking optimal actions. This paper provides a review of recent publications that apply machine learning (ML)-based methods for monitoring power system inertia. It also provides an overview of academic and industrial projects related to ML-based methods for inertia monitoring. Furthermore, the paper explores applications based on ML-based methods and inertia. Lastly, the paper briefly discusses future directions for the development of this research field.