Surface micro-cracks, a common and severe processing defect in CNC milling of high-performance components such as aero-engine turbine blades, act as stress concentration sites that drastically reduce the fatigue life of the component and threaten the structural safety of the engine. However, existing non-destructive testing techniques face challenges in being integrated into machine tools to achieve rapid, online, and quantitative in-situ detection of micro-defects on complex curved components. To address this, this paper proposes a novel in-situ detection paradigm based on a triboelectric nanogenerator (TENG), aiming to achieve high-precision quantitative inversion of the geometric parameters of micro-cracks on machined blade surfaces. The influence mechanism of crack width and depth on pulse amplitude and width was explained through systematic experimental research. Furthermore, by constructing multidimensional signal features that integrate time-domain and frequency-domain features, the coupling effects of width and depth can be decoupled. To achieve automated recognition, we constructed a hybrid deep learning model based on CNN-BiLSTM, which can autonomously mine the intrinsic correlation between spatiotemporal features and crack geometry parameters in the original signal, thereby synchronously outputting accurate predictions of width and depth. The experimental results show that the average absolute error of the model in predicting the width and depth of microcracks is as low as 0.0091 mm and 0.0047 mm, respectively, and the coefficient of determination (R 2 ) is higher than 0.989. Furthermore, the potential of this method for identifying defects such as pitting, linear cracks, and network cracks has been further confirmed. This study not only confirms the enormous potential of TENG in intelligent manufacturing quality online monitoring, but also provides a solid technical path for the development of the next generation of intelligent integrated detection systems.
Tang et al. (2026) studied this question.