The need for an effective and efficient maintenance process increases with the level of complexity of modern rotating machinery. This paper introduces a methodology for transforming raw vibration signals into adequate inputs for machine learning classification algorithms in order to identify present faults in rotating machinery. It complements a previous study by the same authors, which covers the processing of vibrational signals by determining the optimal sampling frequency and using appropriate filters for the raw data. The first part of this study covers feature extraction using time and frequency-domain techniques and correlation matrices are plotted to determine which extracted features are significantly connected and what is the level of their correlation. The study continues with the use of Neighborhood Component Analysis (NCA) where weight factors of the features are calculated in terms of recognizing the present rotating machinery faults. Only the ones with the highest level of importance have been used as input for the classification algorithms. The MATLAB Add-in Classification Learner has been used for training and testing various classification algorithms. K-nearest neighbors classifier (KNN), Support vector machines (SVM), and Wide Neural Network (NN) showed the highest accuracy in distinguishing ten different fault conditions. For this case study, the MaFaulda vibration dataset has been used and ten operating conditions have been considered: normal, imbalance, horizontal misalignment, vertical misalignment, and three faults in the underhang bearing and the overhang bearing.

This study proposes a supervised machine learning framework for vibration-based fault diagnosis of rotating machinery using integrated data-driven and physics-informed feature sets. A dataset acquired under variable load and multiple operating conditions was used for model training. Parallel signal processing techniques were applied to capture fault-related information across multiple frequency bands including time-domain analysis, frequency-domain analysis, baseband analysis, and envelope analysis. From the corresponding signal representations, statistical, spectral, and physics-based features associated with characteristic fault frequencies were extracted and combined into integrated feature sets. The diagnostic performance of models trained using purely data-driven features was systematically compared with models incorporating integrated data-driven and physics-informed features. Support Vector Machine, Random Forests, Gradient Boosting, and an ensemble classifier were evaluated using accuracy, precision, recall, and F1-score metrics. The proposed framework employs a two-layer classification strategy, where the first layer performs multiclass fault identification, while the second layer evaluates the presence of imbalance as a coexisting fault. In addition, the influence of different feature groups as well as individual measurement axes and their combinations on diagnostic performance were analyzed. Validation using a new dataset measured in laboratory conditions confirmed the robustness and generalization capability of the proposed diagnostic framework.

Urban noise is a major environmental concern that affects public health and quality of life, demanding new approaches beyond conventional noise level monitoring. This study investigates the development of an AI-driven Acoustic Event Detection and Classification (AED/C) system designed for urban sound recognition and its integration into smart city application. Using the UrbanSound8K dataset, five acoustic parameters—Mel Frequency Cepstral Coefficients (MFCC), Mel Spectrogram (MS), Spectral Contrast (SC), Tonal Centroid (TC), and Chromagram (Ch)—were mathematically modeled and applied to feature extraction. Their combinations were tested with three classical machine learning algorithms: Support Vector Machines (SVM), Random Forest (RF), Naive Bayes (NB) and a deep learning approach, i.e., Convolutional Neural Networks (CNN). A total of 52 models with the three ML algorithms were analyzed along with 4 models with CNN. The MFCC-based CNN models showed the highest accuracy, achieving up to 92.68% on test data. This achieved accuracy represents approximately +2% improvement compared to prior CNN-based approaches reported in similar studies. Additionally, the number of trained models, 56 in total, exceeds those presented in comparable research, ensuring more robust performance validation and statistical reliability. Real-time validation confirmed the applicability for IoT devices, and a low-cost wireless sensor unit (WSU) was developed with fog and cloud computing for scalable data processing. The constructed WSU demonstrates a cost reduction of at least four times compared to previously developed units, while maintaining good performance, enabling broader deployment potential in smart city applications. The findings demonstrate the potential of AI-based AED/C systems for continuous, source-specific noise classification, supporting sustainable urban planning and improved environmental management in smart cities.

The increasing demand for autonomous wireless sensors in Internet of Things (IoT) applications has intensified research on vibration energy harvesting, particularly in the low-frequency range where ambient vibrations are most prevalent. However, most vibration energy harvesters operate efficiently only at a single resonance mode, resulting in a narrow operational bandwidth and pronounced performance degradation under frequency detuning. To address this limitation, this paper proposes a multimodal hybrid piezoelectric–electromagnetic vibration energy harvester that exploits both the first and second resonance modes of a cantilever-based structure to achieve broadband low-frequency operation. The design is guided by the complementary utilization of strain-dominated and velocity-dominated regions associated with different vibration modes. Numerical modeling and finite element simulations are employed to investigate the influence of mass distribution, deformation characteristics, and relative velocity on energy conversion performance. A secondary cantilever carrying the electromagnetic coil is introduced to enhance the relative motion between the coil and the magnetic field, thereby extending the effective operational bandwidth. The experimental results demonstrate increased harvested power, improved energy conversion efficiency, and a significantly broadened effective frequency range compared to conventional single-mode piezoelectric and electromagnetic energy harvesters.

The concept of energy harvesting systems allows sustainable energy sources, such as mechanical energy, to be harvested from the environment. The feature of piezoelectric material to generate voltage as a result of its deformation characterizes them as an energy harvesting tool, due to their robustness and efficiency characteristics. This paper shows numerical modeling of an energy harvesting cantilever beam using the Euler-Bernoulli method. The FEM simulations of the cantilever beam have been created in order to determine the optimal position of the piezoelectric transducer along the beam. Furthermore, these conclusions have been used to develop the mathematical model in MATLAB in order to investigate the effect of the geometry characteristics of the beam and the piezoelectric transducer on the output parameters. The impact of the dimensions and effective area of the piezoelectric transducer and its location on the cantilever beam as well as the effects of the dimensions of the beam have been studied in order to obtain an optimal energy harvesting model in terms of its efficiency. Once the modeling phase has been completed, the output results regarding the generated voltage and power from the energy harvesting system have been compared and models have been validated.

This study proposes a comprehensive framework for automated diagnostics of rotating machinery using vibration signals. The methodology integrates the stages of dataset selection and preparation, signal processing, feature extraction, dimensionality reduction and training and testing of supervised machine learning algorithms. The preparation of the dataset involves proposing a technique for addressing data imbalance, normalisation of the data amplitude, adjusting the sampling frequency, and calculating the actual rotational frequency. Then, appropriate signal filters are designed and applied, and a wide range of statistical and spectral features is extracted in the time and frequency domain. To address the challenges of high-dimensional feature spaces and allow transparency in identifying features with the highest level of importance in recognising machinery faults, a group-wise dimensionality reduction technique, Sequential Selection Algorithm (SSA), is proposed. The applicability of the proposed SSA technique is validated by comparing its performance with that of commonly used dimensionality reduction techniques, by testing its compatibility with supervised machine learning algorithms, and by analyzing its efficiency. In that sense, a set of models combining four dimensionality reduction techniques and four supervised machine learning algorithms is trained and tested, and their accuracy is compared.