函数链神经网络在火箭大喷管加工中的应用

Application of functional link neural networks to machining rocket jet

函数链神经网络ＦＬＮＮ（Ｆｕｎｃｔｉｏｎａｌ－ＬｉｎｋＮｅｕｒａｌＮｅｔｗｏｒｋｓ）是一种引入非线性扩展函数的单层网络，具有很好的非线性逼近能力．证明了ＦＬＮＮ拟合ｐ个样本的充要条件，并将ＦＬＮＮ应用于火箭大喷管加工中的壁厚数据拟合；解决了原来用三次样条插值拟合时，分段插值函数及求解参数太多等问题，使加工中的切深实时补偿控制简便可行。

Functional link neural network (FLNN) is single layer network in combination with nonlinear expanding functions and has a good ability of nonlinear approach. The condition of FLNN fitting p samples has been proved. Using FLNN in plate thickness curve fitting of machining rocket jet makes on line cutting depth compensation control easy and feasible and ensures the machining precision. Thus , it can overcome the drawbacks of too many interpolating functions and coefficients which need to be solved in using spline curve.