GA-CRBF网络板形识别的DSP实现

Implementing Flatness Recognition Based on GA-CRBF Network by DSP

针对传统RBF网络板形模式识别方法存在抗干扰能力差、识别精度有限以及缺乏处理不确定性信息的能力等问题,将云模型引入RBF神经网络,提出一种新型板形识别模型。MATALB仿真结果表明:新型GA-CRBF网络正确识别出板形缺陷,识别精度比传统的RBF网络提升73%,抗干扰性也提升了83%。将GA-CRBF网络写入DSP芯片中运行后,正确识别出缺陷板形,验证了其工程应用的可行性,为神经网络推广应用到实际工程中提供了依据。

In view of poor anti-interference ability,limited recognition accuracy and poor capacity for dealing with uncertainties of the conventional RBF neural network,a cloud model was introduced into the Network to establish a novel flatness recognition model. MATALB simulation results showed that new GA-CRBF network could correctly identify flatness defect,with the recognition accuracy and anti-interference capacity increased by 73% and 83% respectively,compared with the traditional RBF network. With GA-CRBF network written into the DSP chip,it can correctly identify defective flatness after running,verifying the feasibility of its application into engineering,which lays foundation for an application of neural network into practical engineering.