Recently, transfer learning technology has provided valuable solutions to problems that are present in machinery with industrial applications. Through the use of transfer learning, basic diagnostic problems have been well addressed, especially in scenarios in which the training and test data are from different distributions. However, there are scenarios that require further consideration, such as partial fault diagnosis. In this paper, a deep learning-based fault diagnosis methodology is proposed to address the partial fault diagnosis problem, in which the data from the unsupervised target domain represents a category subspace of the full machine-state-label space. Specifically, a domain adaptation with adversarial learning schemes is proposed to achieve partial domain adaptation. The experimental results on an electromechanical test bench suggest that the proposed approach offers a practical solution to this partial fault diagnosis problem.

This paper shows the preliminary results of a proposed methodology for outer race fault diagnosis of contaminated rolling bearings. The proposed method is based on a low-cost thermographic sensor. This sensor allows for the acquisition of thermographic images, which are subsequently processed by means of statistical and non-statistical indicators. The obtained results shows a correct fault classification of 4 condition states as follows: healthy rolling bearing, outer race fault + 1 g of magnetite-contaminated bearing, outer race fault + 2 g of magnetite-contaminated bearing, and outer race fault + 3 g of magnetite-contaminated bearing

Critical situations such as the current pandemic produced by the Sars-Cov-2 (COVID-19) are leading to the incorporation of new technologies, such as virtual reality (VR), as a part of the teaching and training methods that facilitates the teaching task in contrast with the traditional learning methods that have been used in the academia. In this regard, in this work is proposed the design and development of a VR-based tool that is applied in the teaching and training of condition-based maintenance strategies in induction motors. The design of this tool is performed under Oculus Quest and includes the development of the 3D virtual environments, the creation of the VR learning experience, and finally the necessary adjustments for the VR application to render smoothly on standalone devices. The usability of this experience tool has been experimentally validated with different students by giving to the students practical knowledge but also a learning experience.

In the modern industry framework, the application of condition monitoring schemes over electromechanical systems is being subjected to demanding requirements. Currently, the massive digitalization of industrial assets allows the investigation towards multiple monitoring strategies capable of emphasize deviations over the nominal system operation. However, the most prominent techniques, such as Machine Learning, present great challenges in complex systems. In this regard, the proposed study presents the analysis of the diagnostic capabilities resulting from the classical approaches based on machine learning facing to complex electromechanical systems that implies a working environment subject to different operation condition, configurations with multiple components and the presence of faults of different nature (mechanical, electrical, electromagnetic), under isolated or combined scenarios. Discriminative feature extraction capabilities and classification accuracy will be analyzed as performance measures.

Condition monitoring applied to power quality involves several techniques and procedures for the assessment of the electrical signal. Data-driven approaches are the most common methodologies supported on data and signal processing procedures. Electrical systems in factory automation become more complex with the increase of multiple load profiles connected, and unexpected electrical events can occur causing the appearance of power quality disturbances. However, emerging technologies in the techniques related to the detection and identification of power quality disturbances are analyzed and compared according to the complexity of the current electrical system, that is, including simple and combined disturbances. These new technologies allow developing more cyber-physical systems to process the new methodologies for condition monitoring. Thus, in this study, a deep learning-based approach for the identification of power quality disturbances is implemented and their performance analyzed in front of classical disturbances defined by the International standards considered in the related literature.

The detection of uncharacterized events during electromechanical systems operation represents one of the most critical data challenges dealing with condition-based monitoring under the Industry 4.0 framework. Thus, the detection of novelty conditions and the learning of new patterns are considered as mandatory competencies in modern industrial applications. In this regard, this article proposes a novel multifault detection and identification scheme, based on machine learning, information data-fusion, novelty-detection, and incremental learning. First, statistical time-domain features estimated from multiple physical magnitudes acquired from the electrical motor under inspection are fused under a feature-fusion level scheme. Second, a self-organizing map structure is proposed to construct a data-based model of the available conditions of operation. Third, the incremental learning of the condition-based monitoring scheme is performed adding self-organizing structures and optimizing their projections through a linear discriminant analysis. The performance of the proposed scheme is validated under a complete set of experimental scenarios from two different cases of study, and the results compared with a classical approach.

Condition monitoring in electromechanical systems represents, currently, one of the most critical challenges dealing with the advancement and modernization in intelligent manufacturing. In this regard, machine learning based algorithms widely applied in other technological fields are being considered now to face the automatic feature extraction on the electric machine monitoring. In this study, a monitoring scheme is considered for faults detection performance evaluation, where vibrations signal under different fault conditions are acquired. Thus, the common electric machine monitoring framework, that is, a set of features estimated from a limited number of measurements, is considered in front of the three main dimensionality reduction approaches: principal component analysis, linear discriminant analysis and auto-encoder based. Performance of the corresponding approaches are studied and discussed experimentally. It is revealed that, although scheme based on auto-encoder provides enhanced diagnosis results, it is still necessary to carry out a detailed study on the automatic extraction capabilities of important features for the detection of faults.

The demand of electric power supply has increase in the industry due to most of its processes are involved with the use of electrical equipment and machines. Electric power generation due to integration of new energy sources has become an important area of continuous development in which Power Quality (PQ) problems must be faced. In this paper is proposed a condition monitoring strategy based on the estimation of statistical time domain-based features and Linear Discriminant Analysis to identify different PQ disturbances in a cogeneration system. The proposed method is first evaluated with a set of synthetics signals that include different PQ disturbances and then evaluated a real data acquired from a cogeneration system. The final diagnosis outcome is performed by means of a Neural Network. The obtained results shown that the proposed method is suitable for being applied in cogeneration system to identify different PQ disturbances.

This study is focused on the current challenges dealing with electromechanical system monitoring applied in industrial frameworks, that is, the presence of unknown events and the limitation to the nominal healthy condition as starting knowledge. Thus, an industrial machinery condition monitoring methodology based on novelty detection and classification is proposed in this study. The methodology is divided in three main stages. First, a dedicated feature calculation and reduction over each available physical magnitude. Second, an ensemble structure of novelty detection models based on one-class support vector machines to identify not previously considered events. Third, a diagnosis model supported by a feature fusion scheme in order to reach high fault classification capabilities. The effectiveness of the fault detection and identification methodology has been compared with classical single model approach, and verified by experimental results obtained from an electromechanical machine.

A great deal of investigations are being carried out towards the effective implementation of the 4.0 Industry new paradigm. Indeed, most of the machinery involved in industrial processes are intended to be digitalized aiming to obtain enhanced information to be used for an optimized operation of the whole manufacturing process. In this regard, condition monitoring strategies are being also reconsidered to include improved performances and functionalities. Thus, the contribution of this research work lies in the proposal of an incremental learning framework approach applied to the condition monitoring of electromechanical systems. The proposed strategy is divided in three main steps, first, different available physical magnitudes are characterized through the calculation of a set of statistical-time based features. Second, a modelling of the considered conditions is performed by means of self-organizing maps in order to preserve the topology of the data; and finally, a novelty detection is carried out by a comparison among the quantization error value achieved in the data modelling for each of the considered conditions. The effectiveness of the proposed novelty fault identification condition monitoring methodology is proved by means of the evaluation of a complete experimental database acquired during the continuous working conditions of an electromechanical system.

The procedures for condition monitoring of electromechanical systems are undergoing a reformulation, mainly, due to the current thermographic affordability of infrared cameras to be incorporated in industrial applications. However, high-performing multifault data-driven methodologies must be investigated in order to infer reliable condition information from the thermal distribution of not only electrical motors but also of shafts and couplings. To address this issue, a novel thermography-based methodology is proposed. First, the infrared capture is processed to obtain a thermographic residual image of the kinematic chain. Second, the thermal distribution of the image's regions of interest is characterized by means of statistical features. Finally, a distributed self-organizing map structure is used to model the nominal thermal distribution to subsequently perform a fault detection and identification. The method provides a reliability quantification of the resulting condition assessment in order to avoid misclassifications and identify the actual fault root-causes. The performance and the effectiveness of the proposed methodology is validated experimentally and compared with the classical maximum temperature gradient procedure.

Gearboxes are important components in industrial applications and its condition monitoring is relevant in industries for reducing costs and minimizing maintenance downtime. Diagnosing wear in gearboxes saves time to prepare appropriate corrective actions, and to ensure that the system does not deteriorate critically. Nowadays, for condition monitoring in gearboxes, vibration analysis is commonly used due to its high reliability. In this work, a reliable methodology for diagnosing different levels of wear in a gearbox through vibration signals, and supported by a theoretical model, is proposed. The theoretical model is based on calculating the characteristic frequencies of the gearbox, with the aim for locating the spectral components of the faults in the vibration signal. Experimentation is done to a healthy gearbox and three wear levels. Results show the reliability of this method that makes it suitable to be used in diagnosing industrial machinery such as in automotive manufacturing applications.