This study presents a novel fault diagnosis approach for milling machines by combining Acoustic Emission (AE) with advanced deep learning techniques. AE data is collected under various fault conditions, including bearing, tool, gear faults, and normal operations. The data is converted into 2D representations using Short-Time Fourier Transform (STFT) to capture the non-stationary behavior of the signals. Sobel edge filtering enhances feature clarity and reduces noise, emphasizing key signal patterns. For feature extraction, Convolutional Neural Networks (CNN) are employed to capture local features, while Convolutional Autoencoders (CAE) are used to extract global features. These complementary features are combined into a “Unified Feature Matrix” and fed into an Artificial Neural Network (ANN) for fault classification. The proposed method effectively identifies and classifies fault conditions with high accuracy, demonstrating its ability to detect bearing faults, tool wear, gear damage, and normal operations while overcoming noise and computational complexity challenges.