Search

Induction machines (IMs) play a critical role in various industrial processes but are susceptible to degenerative failures, such as broken rotor bars. Effective diagnostic techniques are essential in addressing these issues. In this study, we propose the utilization of convolutional neural networks (CNNs) for detection of broken rotor bars. To accomplish this, we generated a dataset comprising current samples versus angular position using finite element method magnetics (FEMM) software for a squirrel-cage rotor with 28 bars, including scenarios with 0 to 6 broken bars at every possible relative position. The dataset consists of a total of 16,050 samples per motor. We evaluated the performance of six different CNN architectures, namely Inception V4, NasNETMobile, ResNET152, SeNET154, VGG16, and VGG19. Our automatic classification system demonstrated an impressive 99% accuracy in detecting broken rotor bars, with VGG19 performing exceptionally well. Specifically, VGG19 exhibited high accuracy, precision, recall, and F1-Score, with values approaching 0.994 and 0.998. Notably, VGG19 exhibited crucial activations in its feature maps, particularly after domain-specific training, highlighting its effectiveness in fault detection. Comparing CNN architectures assists in selecting the most suitable one for this application based on processing time, effectiveness, and training losses. This research suggests that deep learning can detect broken bars in induction machines with accuracy comparable to that of traditional methods by analyzing current signals using CNNs. [ABSTRACT FROM AUTHOR]

Induction machines (IMs) are essential components of many industrial installations and, therefore, their faults must be detected early. Fault detection using current spectrum analysis is attracting an increasing interest as a condition-based monitoring (CBM) technique. However, its use to detect rotor asymmetries in high-power IMs, which operate at very low slip, is particularly challenging, due to the closeness of the characteristic fault harmonics to the fundamental component, separated only a few MHz. Their reliable detection in harsh industrial environments requires a very high spectral resolution, that is, long acquisition times and a huge number of current samples, which hinders its implementation on embedded, online devices with limited computing resources. To address this problem, this article presents a novel combination of diagnostic techniques, the use of the rectified current as diagnostic signal, and the Goertzel algorithm as signal processing tool. This unique combination allows for an optimized implementation of the Goertzel algorithm, which provides a high spectral resolution in the full load range of the machine, with a low computational cost and a negligible memory footprint. This proposal is validated experimentally with the fault diagnosis of a high-power medium-voltage industrial motor. [ABSTRACT FROM AUTHOR]

Motor current signature analysis has become a widespread fault diagnosis technique for induction machines (IMs), because it is noninvasive and requires low resources of hardware (a current sensor) and software (a fast Fourier transform). Nevertheless, its industrial application faces practical problems. One of its most challenging scenarios is the detection of broken bars in IMs working at very low slip, like large machines with a very small rated slip, or unloaded induction motors in off-line tests. In these cases, the leakage of the main supply component can hide the fault harmonics, even with a severe fault. Diverse solutions to this problem have been proposed, such as the use of smoothing windows, advanced spectral estimators, or the removal of the supply component. Nevertheless, these methods modify the spectral content of the current signal or add a high computational burden. In this work, a new approach is proposed, based on the analysis of the current with a very fine spectrum, obtained via simple zero padding, followed by the extraction of a practically leakage-free conventional, coarse spectrum. The method is experimentally validated by the diagnosis of a broken bar fault in a 3.15-MW induction motor. [ABSTRACT FROM AUTHOR]

Fault diagnosis of induction motors through the analysis of the stator current is increasingly being used in maintenance systems, because it is non-invasive and has low requirements of hardware and software. Nevertheless, its industrial application faces some practical limitations. In particular, the detection of fault harmonics that are very close to the fundamental component is challenging, as in large induction motors working at very low slip, because the leakage of the fundamental can hide the fault components until the damage is severe. Several methods have been proposed to alleviate this problem, although all of them increase noticeably the complexity of the diagnostic system. In this paper, a novel method is proposed, based on the analysis of the rectified motor current. It is shown that its spectrum contains the same fault harmonics as the spectrum of the original current signal, but with a much lower frequency, and free from the fundamental component leakage. Besides, the proposed method is very easy to implement, either by software, using the absolute value of the current samples, or by hardware, using a simple rectifier. The proposed approach is presented theoretically and validated experimentally with the detection of a broken bars fault of a large induction motor. [ABSTRACT FROM AUTHOR]

This paper introduces a new approach for improving the fault diagnosis in induction motors under time-varying conditions. A significant amount of published approaches in this field rely on representing the stator current in the time-frequency domain, and identifying the characteristic signatures that each type of fault generates in this domain. However, time-frequency transforms produce three-dimensional (3-D) representations, very costly in terms of storage and processing resources. Moreover, the identification and evaluation of the fault components in the time-frequency plane requires a skilled staff or advanced pattern detection algorithms. The proposed methodology solves these problem by transforming the complex 3-D spectrograms supplied by time-frequency tools into simple x-\rm y graphs, similar to conventional Fourier spectra. These graphs display a unique pattern for each type of fault, even under supply or load time-varying conditions, making easy and reliable the diagnostic decision even for nonskilled staff. Moreover, the resulting patterns can be condensed in a very small dataset, reducing greatly the storage or transmission requirements regarding to conventional spectrograms. The proposed method is an extension to nonstationary conditions of the harmonic order tracking approach. It is introduced theoretically and validated experimentally by using the commercial induction motors feed through electronic converters. [ABSTRACT FROM AUTHOR]

Induction machines drive many industrial processes and their unexpected failure can cause heavy production losses. The analysis of the current spectrum can identify online the characteristic fault signatures at an early stage, avoiding unexpected breakdowns. Nevertheless, frequency domain analysis requires stable working conditions, which is not the case for wind generators, motors driving varying loads, and so forth. In these cases, an analysis in the time-frequency domain—such as a spectrogram—is required for detecting faults signatures. The spectrogram is built using the short time Fourier transform, but its resolution depends critically on the time window used to generate it—short windows provide good time resolution but poor frequency resolution, just the opposite than long windows. Therefore, the window must be adapted at each time to the shape of the expected fault harmonics, by highly skilled maintenance personnel. In this paper this problem is solved with the design of a new multi-band window, which generates simultaneously many different narrow-band current spectrograms and combines them into as single, high resolution one, without the need of manual adjustments. The proposed method is validated with the diagnosis of bar breakages during the start-up of a commercial induction motor. [ABSTRACT FROM AUTHOR]

Fault diagnosis of rotor asymmetries of induction machines (IMs) using the stator current relies on the detection of the characteristic signatures of the fault harmonics in the current spectrum. In some scenarios, such as large induction machines running at a very low slip, or unloaded machines tested offline, this technique may fail. In these scenarios, the fault harmonics are very close to the frequency of the fundamental component, and have a low amplitude, so that they may remain undetected, buried under the fundamental's leakage, until the damage is severe. To avoid false positives, a proven approach is to search for the fault harmonics in the current envelope, instead of the current itself, because in this case the spectrum is free from the leakage of the fundamental. Besides, the fault harmonics appear at a very low frequency. Nevertheless, building the current spectrum is costly in terms of computing complexity, as in the case of the Hilbert transform, or hardware resources, as in the need for simultaneously sampling three stator currents in the case of the extended current Park's vector approach (EPVA). In this work, a novel method is proposed to avoid this problem. It is based on sampling a phase current just twice per current cycle, with a fixed delay with respect to its zero crossings. It is shown that the spectrum of this reduced set of current samples contains the same fault harmonics as the spectrum of the full-length current envelope, despite using a minimal amount of computing resources. The proposed approach is cost-effective, because the computational requirements for building the current envelope are reduced to less than 1 % of those required by other conventional methods, in terms of storage and computing time. In this way, it can be implemented with low-cost embedded devices for on-line fault diagnosis. The proposed approach is introduced theoretically and validated experimentally, using a commercial induction motor with a broken bar under different load and supply conditions. Besides, the proposed approach has been implemented on a low-cost embedded device, which can be accessed on-line for remote fault diagnosis. [ABSTRACT FROM AUTHOR]