The optimization of membership function parameters in Fuzzy-based systems is difficult because of complex interdependencies and geometric constraints. Traditional crossover operators like Blend Crossover (BLX- \(\alpha \) ) and Simulated Binary Crossover (SBX) were designed for unconstrained optimization and they cannot use effectively the parameter correlation in Fuzzy systems. In this paper, we consider parallelotope-shaped Blend Crossover (psBLX) for membership function parameter optimization. We conducted 30 independent runs on a Mamdani fuzzy inference system with 2,700 samples and show that psBLX with \(\beta = 1.1\) gives statistically significant performance improvements of up to 22.69% in mean absolute error compared to SBX. Friedman tests and Nemenyi post-hoc analysis confirm the practical significance of these improvements ( \(p < 0.001\) ). The results show that \(\beta > 1.0\) provides optimal exploration characteristics for membership function optimization, improving both theoretical understanding of geometric crossover mechanisms and practical Fuzzy system design methods.