针对数控加工过程的时变性、非线性和不确定性，甚至无法用精确的数学模型加以描述，传统的PID控制显得无能为力，无法在现场环境下根据外部干扰和随机因素实时动态调整CNC中预先设定的切削参数而影响工作效率和产品加工质量，限制了CNC向多变量智能化控制方向发展，已不适应日益复杂的制造过程，所以本文采用神经网络自适应控制理论来对加工过程进行实时监控，自动调节控制器参数，消除复杂因素和不确定性因素的影响，实现加工过程智能控制，提高加工效率。 目前对监控过程采取的诸多算法基本思路仍为自校正控制和模型参考自适应控制，均离不开系统的数学模型，因而其应用受到限制，其主要困难在于加工过程模型的建立和实时优化策略制定。基于神经网络的自适应控制特别适合于非线性时变系统、无法精确建模系统以及操作存在着不确定性的系统，具有较高的智能水平。考虑到BP神经网络具有自学习和对任意非线性函数的万能逼近能力，所以本文提出神经网络建模和优化控制相结合的数控铣削控制系统，在加工过程中不断调整进给速度，从而提高生产率，具有重要的理论意义和现实意义。 本课题以深圳职业技术学院工业中心协鸿HV—35S立式加工中心为研究对象，在充分消化吸收以色列OMAT优铣技术的基础上，对数控铣削系统进行了系统分析和建模，制定了神经网络恒力和恒功率控制方案及控制算法，经模拟仿真表明，该控制方案效果较好，控制精度较高，具有实时性、鲁棒性和稳定性的特点，具有自适应控制功能，实现了数控铣削过程神经网络智能控制。并初步设计出了神经网络智能监控仪，作为三菱数控系统的一个辅助控制器，通过检测主电机电流，调节进给速度，从而实现最佳功率约束自适应控制。

The process of CNC is a complicated, non-linear, unknown and variable dynamic process, and it almost can’t be described by precise mathematical model. So it can’t be dealt with by traditional PID control. The parameters of CNC system can’t be adjusted real time according to the jamming and randomness, which affects efficiency and quality. This method restricts development of CNC to intelligence and can’t adapt to complicated course of production. So a neural network adaptive control algorithm is given in this paper to monitor process of milling. It can adjust parameter of controller automatically in order to eliminate effect of complexity and uncertainty, realize intelligent control and enhance efficiency. Some algorithms used in monitoring CNC process are STR and MRAC presently. These methods, which depend on precise mathematical model, can’t be used widely. The reason is that it is difficult for making model and optimal strategy of system. Adaptive control based on neural network which is especially adapt to nonlinear, various and unknown system has high intelligence. Taking account of self-learning and capacity of approximating unlimitedly, I put forward the NN-NC. Two neural networks are used for system on-line identification and controller. It can adjust the velocity adaptively so as to enhance efficiency. This meaning is important for theory and reality. On the base of learning and studying the OPTIMIL technology of OMAT Company in Israel, mathematical model of milling system is given based on HV-35S in the IC of SZPT, at the same time, an adaptive On-line control constraint system algorithm based on BP neural network is presented. The designed system, which is simulated by computer in CNC milling process, has high adaptive and self-learning ability, and it is suitable for complicated, unknown and variable dynamic process. Simulation has demonstrated that this controller based on BP neural network for adaptive control is stable, reliable and robust. At last, the basic NN controller is designed, which used for helping controller of Mitsubishi CNC system. It can adjust velocity of tool through monitoring the current of main electric machine in order to accomplish optimal constraint adaptive power control in milling process.