In practical engineering systems, accounting for various uncertainties during the design process is paramount. However, reliability analysis in structural engineering often entails substantial computational costs, particularly when dealing with implicit performance functions and scenarios involving very low failure probabilities. This inherent complexity underscores the challenges faced in real-world applications, where efficient and accurate reliability assessments are crucial for ensuring structural integrity and safety. In recent years, the concept of utilizing surrogate models for reliability analysis has garnered significant attention. The approach outlined in this study employs an innovative surrogate framework that concurrently integrates Cascade-forward Neural Networks (CFNN) and Self-Organizing Map (SOM) networks, alongside an optimized kriging model. The final reliability assessment is then determined as a weighted average of the outputs from these integrated models. To comprehensively illustrate the effectiveness of the proposed algorithm, a diverse range of examples are included: five mathematical examples and five engineering examples. Furthermore, a detailed discussion highlights the benefits of this proposed method in comparison to alternative approaches. The results demonstrate the effective performance of the developed methodology. For instance, in the mathematical examples, the minimum improvement observed over other methods is an 81%, coupled with an approximate 0% error in reliability calculation. Similarly, for the engineering examples, a minimum improvement of 47% is achieved over existing methods, with the reliability calculation error remaining low at approximately 1%.