基于Shapley值的古建筑木构件内部缺陷无损检测

Nondestructive Testing of Internal Defect of Ancient Architecture Wood Members Based on Shapley Value

为了准确地获取古建筑木构件内部缺陷状况,采用应力波和阻抗仪2种无损检测方法,对古建筑常用树种榆木和杨木试件在不同缺陷类型、不同内部缺陷面积下进行试验,并基于Shapley值对其内部缺陷进行组合预测.研究结果表明:应力波能快速获取检测截面的二维图形,内部缺陷较小时误差率较大,当内部缺陷面积大于整个检测截面的1/4时应力波检测误差值逐渐减小.阻抗仪对单路径下的内部缺陷能够准确判断缺陷位置及缺陷类型.当木构件内部缺陷较小时阻抗仪检测结果较精确,内部缺陷面积扩大后阻抗仪检测误差变大.应用Shapley值方法,通过分配2种无损检测总误差来确定组合预测模型中应力波和阻抗仪的权重,构建组合预测模型,组合预测误差值均小于10cm2,检测精度高于单一的检测结果,可以为古建筑木构件工程的修缮和加固方案提供有效数据支持.

To master the internal damage of ancient architecture wood members, by employing twonondestructive testing methods including stress wave and impedance instrument, elm and cypress woodspecimen which were different damage types or different internal defect area was tested. Then in thispaper a combinatorial model to forecast internal defects of ancient architecture wood members wereestablished based on Shapley value. Results show that Stress wave can get 2D graphics of detectionsection quickly, and error rate is higher when internal defect is smaller. The stress wave error decreaseswhen internal defect is more than a quarter. Impedance instrument can accurately judge positions anddamage types for internal defects under single path. The impedance instrument test results are moreaccurate when internal defect of wood members is small, internal defects area is proportional toimpedance instrument error when increase. Applying the Shapley value method, the weight of stress wave and impedance instrument in the combination forecast model, was determined by distributing the total combinatorial error, then a combinatorial forecast model was establised. The results suggest that combinatorial forecast error is less than 10, and detection accuracy is higher than that of the single test results. The research can help to detect method from qualitative to quantitative analysis for practical engineering repair reinforcement scheme and provide effective data support.