复杂环境下飞机零部件孔参数测量算法研究

Research on Algorithm for Measuring Hole Parameters of Aircraft Components in Complex Environments

现有孔参数测量方法,难以适用于复杂环境条件测量需要,急需设计新的满足工程实际应用的方法。基于此,设计了识别测量算法,包含了同数控加工机床主轴配置的孔图像获取装置,构建了基于残差卷积网络用于提取孔特征的数学模型,用于精细识别孔径回归数学模型与孔位度量数学模型。采集加工现场零部件孔表面状态图像进行分析,提出方法较主流CH、RH、SH、Hh方法,20组测试样本半径偏差累积分别降低为原来的5/16、1/3、5/16、1/4,累积孔位偏差分别降低为原来的5/39、5/32、5/29、10/83,孔口缺陷平均F1-score值为0.94达到了工业应用0.9的标准。基于实验结果,说明了说提出算法更优的半径拟合与孔位测量能力,满足实际识别检测需要的同时可有效助力航空装置智能化转型升级。

The existing hole parameter measurement methods are difficult to meet the needs of high quality results in complex environmental conditions,so it is urgent to design new methods to meet the practical application of engineering.Based on this,an algorithm was designed,including a hole image acquisition device configured with the same CNC machining tool spindle.A mathematical model based on residual convolution network was constructed for extracting hole features,and a mathematical model for fine recognition of pore size regression and hole position measurement was constructed.Compared with the mainstream CH,RH,SH and Hh methods,the proposed method reduces the cumulative deviation of 20 groups of test sample radius to 5/16,1/3,5/16,1/4,and the cumulative deviation of hole location to 5/39,5/32,5/29,and 10/83,respectively.The average F1-score of orifice defects was 0.94,which reached the standard of 0.9 for industrial applications.Based on the experimental results,it is shown that the proposed algorithm has better radius fitting and hole position measurement ability,which can meet the needs of actual identification and detection and effectively help the intelligent transformation and upgrading of aviation devices.