【摘要】 在工业过程参数优化问题中,由于测量噪声和过程的随机干扰,测量所得的性能曲面往往呈现多极值点的形态,但其中真正的极值点只有一个.传统的方法是在建模(例如神经网络)的过程中自然消除掉部分高频噪声.首先介绍了一种寻优性能指标———邻域平均性能准则,这种准则不是仅以性能曲面上某点的函数值作为性能指标,而是以该点及其周围各点函数的加权平均值作为指标.然后,把该性能指标与一种新的寻优方法———粒子群优化法相结合,给出了混有噪声的多维多极值点函数的寻优算法.一、二维函数的仿真数据初步表明了这种新方法是可行的.更多还原

【Abstract】 In optimization of the industrial process parameters,because of the random interferences of the measured noises and process,the measured performance surface often appears multi-poles shape.But there is only one real pole among them.The traditional method is to dispel some high-frequency noises naturally in modeling(for example neural network).Hence,one kind sought excellent performance index,the(average) performance rule,was put forward, which not only regards one function value on the performance surface as the performance index,but also the weighted average of the point function value and the points around as the performance index.The proposed performance index was integrated with a new seeking(optimization) method,particle swarm optimization. A seeking optimization algorithm of multi-dimensional extreme point mixed with the noises was presented.The simulation results of one-and two-dimensional functions show that the new method is feasible.更多还原