【摘要】 根据离散小波框架理论 ,在有限维Hilbert函数空间中 ,给出了一种基于正交投影算子的小波神经网络构造学习算法 ,并讨论了该算法的几何速度稳定收敛性。它从能量信息角度保证了网络构造的最优化 ,同时体现出信号的时频域特性 ,较好地解决了网络的优化设计问题。通过在某歼击机运动系统故障检测中的仿真应用 ,验证了该算法的优越性能更多还原

【Abstract】 An algorithm based on orthogonal projection operator was proposed to construct the wave-net with respect to the frames theory of discrete wavelet transforms. Its geometrical rate convergence is demonstrated in the finite Hilbert functional space. The algorithm optimizes the constructional process in view of energy distribution information while remaining the signal’s local characteristics in time-frequency domains. An application on fault detection for a fighter verifies its correctness.更多还原