Deep Learning-Based Predictive Vehicle Fault Detection Using Simulated Multi-Sensor Timeseries Data A.A. Anumol, P. Vijayalakshmi, A. Kingsly Jabakumar, T. Arun, S. Prabhavathy, V. Nithya Proceedings of 2nd International Conference on Visual Analytics and Data Visualization Icvadv 2026, 2026 Conventional vehicle health monitoring still depends on hardware-heavy threshold rules that react only after a fault appears. This work presents a software-only predictive maintenance pipeline that combines convolutional neural networks, bidirectional LSTMs and an attention layer to detect and classify vehicle faults from simulated multi-sensor time-series data. The system runs entirely in Google Colab and reaches <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 4 - 9 6 \%}$</tex> classification accuracy, while its remaining useful life head attains a mean absolute error of about <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\pm \mathbf{4. 5}$</tex> cycles. Six deep learning models are trained and compared: three baselines (Bi-LSTM, GRU, 1D-CNN) and three hybrid variants (CNN-LSTM, CNN-BiLSTM, CNN-BiLSTM-Attention). The hybrid CNN-BiLSTM-Attention model improves accuracy over the best baseline by roughly <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$5-6 \%$</tex> and also exposes interpretable attention maps. All models operate on synthetic temperature, vibration and current signals generated for four operating states: Normal, Overheating, Mechanical Fault and Electrical Fault. The best model offers real-time inference of around 50 ms per sample with a footprint of about 1.5 MB, which makes it practical for Industry 4.0 deployments and easy to use in teaching and research without extra hardware.
Diabetic Retinopathy Classification using Hybrid Optimized U-Net and Improved ResNet-18 with MISH Activation S. Nithyapriya, S. Yuvalatha, S. Prabhavathy, T. Arun, S.K Muthusundar, Antonidoss A 2024 International Conference on Smart Technologies for Sustainable Development Goals Icstsdg 2024, 2024 Diabetic Retinopathy (DR) refers to damage to the retina caused by diabetes, which can cause visual impairments or potentially lead to blindness. The process of manually identifying diabetic retinopathy is slow and can easily be affected by human mistakes because of the eye's complex anatomy. This paper aims to determine the optimal model for accurately staging diabetic retinopathy (DR) across five DR categories. The proposed system involves image pre-processing, segmentation and classification. A well-structured preprocessing framework was implemented, integrating Median Filtering technique to decrease noise and Gamma Correction for improved image quality. Data augmentation is performed through methods like flipping, cropping rotating, and translation of fundus images. In this study, a hybrid optimization strategy is proposed by combining the Adam optimizer with Simulated Annealing (SA) and Cosine Annealing for training a U-Net model for diabetic retinopathy (DR) lesion segmentation. Canny Edge Detection is employed to discover edges correctly within the image. For DR categorisation in this work, enhanced Resnet 18 combined with MISH activation function is used. The suggested model makes use of residual blocks in conjunction with identity and convolutional blocks. Experiments on the APTOS dataset show that the proposed model outperforms state-of-the-art techniques. Accuracy, precision, recall, and F1 score are all improved by the suggested Resnet18 model, which receives scores of 99.45%, 98.64%, 98.63%, and 98.63%, respectively.
Neem Leaf Disease Detection based on Enhanced Leaky Capacitor-Fired Neuron (LCFN) Model Anandhi Kathiresan, Sahana Shetty, S M. Ramesh, T. Arun, Praveen Kumar R, Glory E Proceedings of 9th International Conference on Science Technology Engineering and Mathematics the Role of Emerging Technologies in Digital Transformation Iconstem 2024, 2024 In herbal products manufacturing field, Neem plays a vital role in India. Abiotic and biotic agents affects the leaves and growth of Neem tree. Viruses, fungus and bacteria are the well known biotic agents and humidity, temperature and water are the well know abiotic agents. High quality herbal product production and plant growth are reduced due to biotic active ingredients. Bacteria affects the neem leaves. Neem leaf disease symptoms can be easily identified by farmers. Laboratory tests like polymerase chain reaction (PCR) is also used for the identification and it is a costly one. In this work, Leaky Capacitor-Fired Neuron (LCFN) model is proposed to identify the neem diseases. One dimensional time sequences are extracted by LCFN from neem leaves. By combining pixels of neighborhood images, a continuous image sequence is created by LCFN. Optimized features from neem leaves are extracted using LCFN. An ensemble classifier is used to classify the leaves based on extracted features. Better performance are exhibited by LCFN as shown in experimental results.
