Analog circuits within the core of electronic systems can introduce faults in the Circuit Under Test (CUT), which may lead to malfunction or even catastrophic failures. The intermittent and parametric faults are also quite hard to observe because of their random nature and non-reproducibility. In this research, a novel SPOT-IP fault detection model has been proposed to enhance the reliability and safety of an electronic system in the event of analog circuits. Initially, the Equispaced curvelet transform(N-ECT) is used to extract spectral, statistical characteristics from gathered signals containing the frequency response to enhance the localization of features. The proposed method utilizes a Graph Attention Capsule Neural Network (GACNN) that is used to categorize analog circuits as fault-free, parametric fault, and intermittent fault, using extracted features. Additionally, the hyperparameters of the GACNN are optimized using Beluga Whale Optimization to improve detection accuracy. The simulation was validated by the two filters, namely the Sallen-Key bypass filter (SK-BPF), and a Tow-Thomas filter (TTF), and the proposed method is evaluated using accuracy, sensitivity and computational time. The proposed technique exhibits better accuracy for intermittent fault (99.12%) and parametric fault (99.25%) detection which is high compared to the existing techniques.