在航空发动机轮盘损伤容限设计中使用了无损检测可靠性（POD）的关键性指标a_{\mathrm{90}/\mathrm{95}}。针对单尺寸参数POD研究的a_{\mathrm{90}/\mathrm{95}}会造成涡流检测系统能力表征不准确的问题，通过分析轮盘人工刻槽缺陷的长度和深度因素对POD的影响，提出了一种基于双参数表征的模型辅助检测可靠性（MAPOD）模型。该模型假设由双尺寸参数向量x所描述的缺陷的涡流响应幅值服从正态分布，将仿真与涡流试验响应数据代入极大似然估计获取模型参数，进而计算得到表征检测系统的可靠性参数。通过对TC4缺陷产生的涡流信号进行分析，分别比较了缺陷数目对单参数模型与改进模型的影响程度。研究结果表明：相较于单参数POD模型，改进的MAPOD模型能以更少的样本数目、更准确地对无损检测系统能力进行表征，最终利用少量缺陷样本初步评估了航空发动机轮盘涡流检测系统检测能力，进一步为航空发动机检修周期以及损伤容限设计（DT）提供了重要依据。

The key index a_{\mathrm{90}/\mathrm{95}} of Probability of Detection （POD） is used in the design of aero-engine disk damage tolerance. Aiming at the problem that a_{\mathrm{90}/\mathrm{95}} of the POD study on single size parameters causes inaccurate capability representation of the eddy current detection system， this paper proposes a Model-Assisted Probability of Detection （MAPOD） model based on two-parameter representation by analyzing the influence of length and depth factors of manual grooving defects on POD. The model assumes that the eddy current response amplitude of the defect described by the two-dimension parameter vector x follows a normal distribution. The simulation and eddy current test response data are substituted into the maximum likelihood estimation to obtain the model parameters， and the reliability parameters characterizing the detection system are then calculated. After analysis of the eddy current signal generated by the TC4 defect， the influence degree of the number of defects on the single parameter model and the improved model is compared. The results show that the improved MAPOD model can characterize the Nondestructive Testing（NDT） system capability more accurately with fewer samples compared with the single-parameter POD model. Finally， the detection capability of the aero-engine wheel eddy current testing system is evaluated with a small number of defect samples， providing important basis for aero-engine maintenance cycle and DamageTolerance （DT） design.