Automated Anomaly Detection in Tensile Testing Data using Bayesian Hyperparameter Optimization and Convolutional Neural Networks

**Abstract:** This paper introduces a novel approach to automated anomaly detection in tensile testing data using a Bayesian hyperparameter optimization framework coupled with convolutional neural networks (CNNs). Traditional methods rely heavily on manual inspection and threshold-based algorithms, which are often inefficient and prone to human error. Our method leverages the power of CNNs to learn complex patterns from raw stress-strain curves, identifying anomalous behaviors indicative of material defects or experimental errors with significantly improved accuracy and speed. The Bayesian hyperparameter optimization process systematically tunes the CNN architecture to maximize anomaly detection performance, enhancing robustness and adaptability across diverse material types and testing conditions. This system promises to dramatically improve quality control processes in materials research and manufacturing, ultimately leading to enhanced product reliability and reduced waste.