The image-based approach is widely used in Fault Detection (FD) algorithms of mechanical systems. The images are derived from the vibrational signals transformed from the time to time–frequency domain, and they are used to develop a Convolutional Neural Network (CNN) to automate the FD process. Nowadays, images are also obtained from the transformation of vibrational signals from the time domain to Symmetrized Dot Pattern (SDP) coordinates, achieving high CNN testing accuracy. This paper shows a comparison of image-CNN approaches for FD using images obtained from time–frequency transforms and those obtained from the SDP transform as input. The comparison was conducted using experimental data from two publicly available bearing datasets, examining both the accuracy of the CNNs and the computational time required for the vibrational signal transformations. The results show that the SDP-CNN approach achieves the same accuracy as spectrogram-CNN approaches but with a significantly reduced computational time. These results support the future real-time implementation of the SDP-CNN approach for FD in mechanical systems such as bearings. Conflict of Interest Statement The authors declare no conflicts of interest.

The present work proposes a new torque/speed equilibrium point monitoring technique for an aircraft Hybrid Electric Propulsion System (HEPS) through an accelerometric-signal-based approach. Sampled signals were processed using statistical indexes, filtering, and a feature reduction and selection algorithm to train a classification Feedforward Neural Network. A supervised Machine Learning model was developed to classify the HEPS operating modes characterized by an Internal Combustion Engine as a single propulsor or by combining the latter with an Electric Machine used as a motor or a generator. The abnormal changes in the torque/speed equilibrium point were detected by the monitoring index built by computing the Root Mean Square on the value identified by the classifier. The procedure was validated through experimental tests that demonstrated its validity.

In the field of gear Fault Detection, the Symmetrized Dot Pattern (SDP) technique, combined with a Convolutional Neural Network (CNN), is widely used to classify various types of defects. The SDP-CNN combination is used to transform vibration signals and simplify the defect classification process under stationary operating conditions. This work aims to enhance the SDP-CNN combination for detecting incipient defects in gear under variable working conditions. The vibration signals are filtered by Vold-Kalman Filter Multi-Order Tracking to highlight fault characteristics under variable working conditions. Subsequently, the signals are SDP-transformed and following classified by optimized CNN. The new pipeline has been validated on an experimental dataset and compared with the classical one by developing both two and multi-class CNNs. The results showed the applicability of the new pipeline in terms of percentage accuracy and ROC curve compared to the classical approach. Finally, the proposed pipeline was compared with other ML literature techniques using the same dataset.

Sensor fusion algorithms are of fundamental importance for the odometric pose estimation of Differential Drive Wheeled Robots (DDWR) as it can be integrated into other algorithms to localize and map the environment. Different sensors are used for odometric pose estimation and, in some cases, Virtual Sensors (VS) are also used. The purpose of this work is to present a new VS can be integrated into sensor fusion algorithms for estimating the odometric pose of the robot. For this reason, a new VS for real-time estimation of the angular speed of the wheels connected to a DC of a DDWR for indoor applications is presented. The estimated wheel angular speed is obtained by processing the vibrational signals sampled from an onboard Inertial Measurement Unit (IMU) through the Discrete Fourier Transform (DFT). This technique is named Fast Fourier Transform as Wheel Angular Speed Estimator (FFT-WASE). The estimation process, integrated with the IMU, enables the realization of a new VS. The estimated angular speed obtained through the VS-based approach is adopted for making the speed control of the DDWR, as an application to validate the proposed technique. The comparison between the estimated and the measured speed coming from the encoders, the study of the absolute percentage errors and RMSE, the reconstruction of DDWR linear speed and yaw rate, and the stability analysis of the controlled DDWR have proved the validity of VS. Further tests are provided to show the applicability of the same acquired vibrational signals for real-time wheel slip monitoring.

Fault detection (FD) is of primary importance for maintenance of mechanical systems. In recent years, the symmetrized dot pattern (SDP) technique has been increasingly applied in this context. This article introduces a new approach to SDP based on new indices derived from SDP transformed vibrational signals. The indices characterize the density, orientation and curvature of the distribution of ‘snowflake’ diagrams, and they were used as inputs for a feedforward neural network (FNN) for FD of bearings as an application of the new approach. The validity of the technique was demonstrated through its application on two public rolling bearing datasets, thereby substantiating its generalizability across a range of fault types and operating conditions. Results demonstrate high classification accuracy, low false positives rates and low computational costs suitable for real-time implementation. The new SDP and FNN approach was also compared to the classical convolutional neural network-based approach: the new method does not require the images, achieves better performance and has a low computational cost compared to the classical one. Finally, the proposed approach is compared with the most modern techniques demonstrating its validity.