Acoustic emission active channel intruder detection system based on machine learning

Machine learning (ML) based acoustic emission (AE) testing has become one of the mainstream methods for structural health monitoring of materials and structures. In this study, an ML-based approach has been proposed to identify and distinguish AE activities generated on a novel active channel intruder detection system (ACIDS), based on parametric analysis. The proposed AE-based ACIDS is better than conventional intruder detections systems as it is immune to signal loss and environmental noise. The novel design of ACIDS redefines the limits of acoustic detection, discussed briefly in this article. Experiments with different intruder activities subjected on the ACIDS have been used to generate AE signals. The acquired AE signals are then analysed using clustering hits and filtering techniques to distinguish between normal disturbances and potential intruders. Different types of human activities for intruder detection were then distinguished using ML techniques. This paper aims to integrate ML with AE technology to provide real-time warnings and reduce the noise-based triggering and achieving up to 99.9% accuracy in binary classification with optimised models. This system can benefit security applications from an unobtrusive, and reliable IDS solution for perimeter protection, such as a nuclear power plant, a military base, or an international fenceless border fencing.