The objective of this research is to design a deep learning-driven intelligent diagnostic system for electromechanical equipment failures, introducing a novel framework termed OPS-DC-LSTM-AE. The Electromechanical Fault Dataset contains 2,300 labeled samples combining vibration, current, and acoustic features for diagnosing faults under varied operating conditions. Pre-processing is achieved through adaptive noise reduction using wavelet packet decomposition, ensuring cleaner signals. For feature extraction, the FFT is applied to capture both temporal and frequency-domain characteristics. The proposed framework integrates OPS to optimize hyperparameters of the DC-LSTM-AE model. The Convolutional layers capture localized signal features, the LSTM layers model temporal dependencies, while the AutoEncoder ensures dimensionality reduction and latent feature learning. This synergy yields an intelligent system capable of diagnosing both single and compound faults with enhanced precision. Results demonstrate superior accuracy of 98.12% and an F1-score of 98.11% using a Python-implemented model, outperforming conventional deep learning approaches. In conclusion, the OPS-DC-LSTM-AE model establishes a powerful and adaptive diagnostic framework, advancing fault detection capabilities in electromechanical equipment toward intelligent and reliable operations.