【作者】 习贺勋；

【导师】 王晓远；

【作者基本信息】 天津大学 ， 电机与电器， 2004， 硕士

【摘要】 永磁无刷直流电动机是一种性能优越、应用前景广阔的电动机，传统的理论分析及设计方法已比较成熟，它的进一步推广应用，在很大程度上有赖于对控制策略的研究。本文提出了一套基于DSP的全数字无刷直流电动机模糊神经网络双模控制系统，将模糊控制和神经网络分别引入到无刷直流电动机的控制中来。充分利用模糊控制对参数变化不敏感，能够提高系统的快速性的特点，构造适用于调节较大速度偏差的模糊调节器，加快系统的调节速度；由于神经网络既具有非线性映射的能力，可逼近任何线性和非线性模型，又具有自学习、自收敛性，对被控对象无须精确建模，对参数变化有较强的鲁棒性的特点，构造三层BP神经网络调节器，来实现消除稳态偏差的精确控制。以速度偏差率为判断依据，实现模糊和神经网络两种控制模式的切换，使系统在不同速度偏差段快速调整、平滑运行。此外充分利用系统硬件构成的特点，采用适当的PWM输出切换策略，最大限度的抑制逆变桥换相死区；通过换相瞬时转矩公式推导和分析，得出在换相过程中保持导通相功率器件为恒通，即令PWM输出占空比D=1，来抑制定子电感对换相电流影响的控制策略。上述抑制换相死区和采用恒通电压的控制方法，减小了换相引起的转矩波动，使系统电流保持平滑、转矩脉动大幅度减小、系统响应更快、并具有较强的鲁棒性和实时性。在这种设计下，系统不仅能实现更精确的定位和更准确的速度调节，而且可以使无刷直流电动机长期工作在低速、大转矩、频繁起动的状态下。本文选用TMS320LF2407作为微控制器，将系统的参数自调整模糊控制算法，BP神经网络控制算法以及PWM输出，转子位置、速度、相电流检测计算等功能模块编程存储于DSP的E2PROM，实现了对无刷直流电动机的全数字实时控制，并得到了良好的实验结果的结果。更多还原

【Abstract】 Because of its eminent performance, the brushless DC motor (BLDCM) has been broadly applied. The traditional theoretical analysis and designing method have been developed maturely. It can be said that the wider applications of BLDCM depend somewhat upon the further research on its special control method. In this paper, fuzzy control and neural network are adopted. When the gap between reference rotary speed and the feedback is huge, regulator based on the rule of fuzzy control is used, which can make the speed follow instruction very quickly. But this method can’t kill tiny discrepancy. So when the gap is small enough, control strategy will switch to the method based on BP neural network. Since a fuzzy BP neural network dual-mode control mode is applied in this paper, the BLDCM control system which is a non-linear and strong coupled system, can gain features of self-adaptive, insensitive and robust to the change of parameters. Generation of BLDCM’s ripple torque is analyzed in detail. Combined with the characteristic of DSP’s symbol flags, the most suitable way is chosen to diminish the dead region of inverter bridge. Mathematic model of torque is induced. When phase current is switched, keeping the IGBT closed in the circuit can restrain the current delay caused by stator winding’s inductance effect. Using TMS320lf2407 as control unit, program modules of fuzzy BP neural network dual-mode control method, output of PWM, feedback collection and A/D conversion are stored in the E2PROM. Verified by experiments, the all-digital control system based on DSP increases the real-time performance and realizes accurately speed control.更多还原