Automatic Sensor less Rotational Velocity Estimation on Asynchronous Machines for Vibration Analysis

2025-03-01T09:01:14+00:00

Detection of incipient faults in rotating machinery is essential to optimize their availability. Online vibration monitoring is a widely recognized approach to gauge the health of machines such as induction motor driven centrifugal pumps. Common faults in such a set-up can be detected by analysing specific fault-related frequency components in the vibration signal, which are almost exclusively (harmonic) multiples of the rotational speed. However, the exact rotational speed is often unknown, which severely complicates the vibration analysis. Online vibration measurements can yield large amounts of data, which necessitates the automation of data processing. This paper presents a novel method that automatically determines the rotational speed from an individual vibration measurement using minimal prior knowledge of the measured system. The proposed method determines the rotational speed with a specified maximum error of 0.0165 Hz for 90.1 percent of cases in the recent dataset containing over 2300 measurements. This paper demonstrates that the proposed method significantly improves upon a comparable method from the literature, as shown through a benchmark comparison.

Content-based Image Retrieval: Innovations in Feature Extraction and User-centric Design

2025-02-07T04:52:40+00:00

Content-based Image Retrieval (CBIR) is a cutting-edge technology that facilitates the search and retrieval of images from large datasets based on visual content rather than metadata. By analyzing intrinsic features such as color, texture, shape, and spatial relationships, CBIR systems address the limitations of traditional text-based approaches, which rely heavily on subjective and labor-intensive annotations. Recent advancements in machine learning, particularly deep learning, have significantly enhanced the efficiency and accuracy of CBIR systems by automating feature extraction and bridging the semantic gap between low-level image features and high-level user interpretations. CBIR systems leverage feature descriptors and indexing techniques to enable fast and accurate image searches, while Convolutional Neural Networks (CNNs) play a transformative role by learning complex hierarchical representations directly from raw image data. Additionally, pre-trained models and transfer learning have further expanded the capabilities of these systems to handle diverse and large-scale image repositories. Experimental results highlight the effectiveness of various feature extraction techniques and the pivotal role of CNNs in modern CBIR frameworks. The findings underscore the importance of integrating user feedback, relevance feedback mechanisms, and multimodal data for creating more intuitive, personalized, and scalable systems. Future directions include exploring unsupervised and self-supervised learning methods, enhancing cross-modal retrieval capabilities, and addressing ethical considerations such as data privacy and algorithmic biases.

Journal of Advancement in Electronics Signal Processing

Automatic Sensor less Rotational Velocity Estimation on Asynchronous Machines for Vibration Analysis

Content-based Image Retrieval: Innovations in Feature Extraction and User-centric Design