文中首先对模糊系统的两类主要的学习算法：梯度下降和遗传算法方法进行了深入分析，并指出了存在的问题．然后，在此基础上提出了一种针对半梯形和三角形隶属度函数的保证隶属度函数ε完备性和模糊集语义一致性的参数调整方法．并基于上述方法实现了一种新的基于遗传算法并利用梯度下降的快速模糊系统学习算法．最后通过实例进行了模拟，验证了该方法的高效性，以及保证隶属度函数完备性和模糊集合语义一致性的优点 Firstly, two kinds of main learning algorithms of fuzzy system: gradient descent and genetic algorithm are analyzed in depth, and the existing problems are pointed out. Then, based on this, a parameter adjustment method is proposed to ensure the membership function ε-completeness and the fuzzy set semantic consistency for the semi-trapezoidal and triangular membership functions. And based on the above method, a new fast fuzzy system learning algorithm based on genetic algorithm and using gradient descent is implemented. Finally, the simulation is carried out by examples to verify the efficiency of the method and the advantages of ensuring the completeness of the membership function and the semantic consistency of fuzzy sets