基于SSA-GA-BP神经网络的数显千分表非线性误差补偿

Nonlinear Error Compensation of Digital Dial Indicator based on SSA-GA-BP Neural Network

利用数显千分表进行精密测量时,零部件的生产、装配及使用磨损、挤压、碰撞等带来的固有误差与弹性误差严重降低了测量精度。针对此问题,利用遗传算法(genetic algorithm,GA)寻优速度快、精度高、并行搜索能力的优势及麻雀搜索算法(sparrow search algorithm,SSA)的全局寻优性能,优化反向传播(back propagation,BP)神经网络的初始权值、阈值及网络结构等,提出了基于数显千分表测量数据非线性误差补偿的SSA-GA-BP神经网络模型。将其与传统BP神经网络、遗传算法优化的GA-BP神经网络进行比较分析。结果表明:所提出SSA-GA-BP神经网络可使数显千分表的非线性误差由没有补偿前的最大误差5.504μm降低至0.883μm,残差平方和、相对误差和R相关系数具有一定的优越性。

When the digital dial gauge is used for precision measurement,the inherent error and elastic error caused by the production,assembly and use of wear,extrusion and collision of parts seriously reduce the measurement accuracy.Aiming at this problem,using the advantages of genetic algorithm(GA)with fast optimization speed,high precision and parallel search ability,and the global optimization performance of sparrow search algorithm(SSA),the initial weights,thresholds and network structure of back propagation(BP)neural network are optimized,and a SSA-GA-BP neural network model based on nonlinear error compensation of digital dial gauge measurement data is proposed.It is compared with the evaluation indexes of traditional BP neural network and GA-BP neural network optimized by genetic algorithm.The results show that the proposed SSA-GA-BP neural network can reduce the nonlinear error of the digital dial gauge from the maximum error of 5.504μm before compensation to 0.883μm within the measurement range.The evaluation index data has certain advantages,and the feasibility of using the neural network model for fitting compensation is verified.