Enhanced threshold detection method enabling fast and robust primary wave time range estimation

The ultrasonic pulse transmission method is a well-established non-destructive testing technique in industry and research. The analysis of the resulting signal data typically aims to determine the time range, as this allows for estimating the speed of sound, a crucial material property. One of the fundamental methods to estimate the time range is threshold detection. Unfortunately, the detection results are biased due to the shape of the signal and the chosen threshold value. In the existing literature, several approaches to reducing this bias are present. The a posteriori signal processing method presented in this paper is based on a threshold detection method with two thresholds and automatic error correction, as described in the literature. However, it also introduces several modifications to enhance the method's robustness concerning frequency and noise. These mainly consist of threshold value optimisation and systematic error correction by taking into account the signal's shape in the detection zone. The analysis results, based on synthetic and natural signals, clearly demonstrate that the proposed method remarkably reduces the residual detection error while requiring moderate computational effort. Since this method does not involve filtering, noise is the governing factor limiting its application range. A comparison of different method variants with methods adapted from literature shows that the proposed method is fast, robust and outperforms the literature-based methods, especially in the regime of low frequencies and increasingly dominant noise.