检测数据驱动的飞机装配偏差传递与装配公差再设计

There are many unfavorable factors which will result in a large uncertainty of aircraft assembly deviation, such as deformation resilience of thin-walled part, riveting interference, deviation coupling of multi-stage assembly and so on. So, the explicit expression can hardly be built for assembly deviation analysis based on the equation of assembly dimension chain. In addition, the small-batch aircraft development has caused the measured data of assembly deviation to have small-sample, information incompleteness and other characteristics; therefore, it is very difficult to analyze the measured data using the traditional statistical methods. . For this reason, this project takes the small-sample, high-dimension, incomplete information and uncertain aircraft assembly as research object, data mining and reasoning methods are introduced to assembly deviation analysis and its quality control, such as measurement information theory, reasoning under uncertainty, comprehensive inversion and so on. The systematic research on gross error judgment, deviation source diagnosis, deviation accumulation and decompose and assembly quality control is carried out based on small sample data. The mechanism of aircraft product assembly deviation transmission will be revealed from another new perspective, and assembly tolerance will be redesigned based on reliable measured data.. This project will improve the traditional assembly deviation analysis method which is based on assembly dimension chain, and provide a new solution for assembly quality control of aircraft products. Besides, it can provide guidance for assembly quality control of other aeronautics and astronautics products.

薄壁件变形回弹、铆接干涉、多装配阶段耦合等不利因素，导致飞机装配偏差存在较大的不确定度，几乎不能构建基于装配尺寸链方程的装配偏差显式计算表达式。此外，飞机研制批量小又导致偏差检测数据呈现出样本小、信息不完备等特点，难以用传统的统计方法进行分析。. 为此，本项目以高维、小样本、信息不完备的飞机不确定装配为研究对象，将测量信息论、不确定性推理和综合反演等数据挖掘和推理方法引入到装配偏差分析及其质量控制中。围绕小样本数据驱动的关键偏差源诊断、偏差累积与分解、偏差波动区间预测和装配质量控制等开展系统的研究。从另一全新视角去揭示飞机装配偏差传播机制，实现基于可靠实测数据的装配公差再设计。. 通过本项目的研究，可补充和完善基于装配尺寸链的传统装配偏差分析方法，为飞机产品装配质量控制提供新的解决途径，也为其它航空航天产品装配质量控制提供有益指导。

薄壁件变形回弹、铆接干涉、装配测量基准误差等不利因素，导致飞机装配偏差存在较大的不确定性，为提高飞机装配质量，降低装配成本，本项目对装配粗差判定、关键偏差源诊断、装配偏差传递与预测、装配质量综合反演及装配公差再设计等开展了系统研究。将测量信息论、传递熵、证据理论等数据挖掘方法引入到装配偏差传递过程中：①提出了飞机结构件装配粗差判定方法。②诊断出影响结构件装配质量的关键偏差源。③提出了基于复杂网络和加权传递熵的复杂薄壁结构装配偏差传递方法。④建立了融合多源数据的飞机结构件装配偏差协同预测模型。⑤提出了基于证据理论的装配偏差仿真模型综合可信度评价方法。⑥构建了基于支持向量机的装配偏差与装配质量映射数学模型。⑦提出了数据驱动的飞机结构件装配公差再设计方法。本项目的研究成果，能有效挖掘小批量研制中产生的可靠信息，实现装配粗差和关键偏差源的准确识别；通过融合装配单元层和产品层偏差检测数据，实现飞机结构件装配质量的准确预测；在装配偏差与装配质量映射数学模型的基础上，实现装配公差再设计。飞机右上壁板装配应用案例计算分析结果表明：1）与经典的格拉布斯粗差判定方法相比，本项目提出粗差识别方法的识别准确率提高了12.5%；2）与3DCS仿真分析方法相比，本项目提出的关键偏差源诊断方法更能客观的量化偏差源的重要程度。飞机中央翼盒装配应用案例计算分析结果表明：相比于传统的灰色预测方法，本项目提出的装配偏差协同预测方法的预测精度提高了9.5%。飞机中上壁板装配应用案例计算分析结果表明：1）利用本项目提出的装配偏差与装配质量映射方法构建的回归模型的均方根误差为0.029，表明该回归模型具有较高的精度。2）利用本项目提出的装配公差再设计方法对飞机中上壁板进行优化设计，优化前后的综合制造成本得到较大减少。本项目为飞机产品装配偏差诊断和装配公差再设计提供新的解决途径，补充和完善了主要依靠制造经验的传统飞机制造质量控制方法。