Towards an Air-coupled Ultrasonic Sensor Network and Camera for Non-destructive Testing with Integrated Machine Learning

There are basically three different measuring techniques with Guided Ultrasonic Waves (GUW) used for Non-destructive Testing (NDT) with respect to the coupling method: Contact-based surface-coupled transducers, immersion fluid coupled transducers, and air-coupled transducers and microphones. We introduce and demonstrate a fully low-cost air-coupled measuring technique with a 1$ MEMS microphone as a signal receiver and a surface-coupled transducer as a sender. Each sensor is fully integrated with data processing and communication providing a "Smart Sensor Node", performing signal processing, feature extraction and Machine Learning locally for feature prediction. Initially, only one microphone is used in a spatial scanning mode. In a next step an array of such MEMS microphones is created, e.g. a 3 × 3 matrix network (or more nodes), towards an spatially extended Ultrasonic camera. To avoid a separation of the microphones from the analog-digital conversion (ADC) processing units (with a bunch of cables) we couple the microphones directly to a microcontroller placed near by the microphone with a 1:1 node mapping. Each node uses a low-cost 32-Bits STM32 ARM microcontroller equipped with ADC and communication controllers. All nodes are connected in a mesh-like grid creating a distributed communicating network. We will deploy and evaluate the first mock-up prototype for fast and real-time hidden defect detection in laminate structures (e.g., Fibre-Metal Laminates). This work addresses ultrasonic monitoring as well as distributed computing and distributed algorithms. Compared with conventional transducer arrays coupled to a single measuring and signal processing unit, our approach provides a much higher freedom of design and there is basically no limit on scalability, especially regarding spatially large extended sensor networks. The typical spacing of sensor nodes (microphones) is about 2-5 cm. The aim is to provide an active measuring system that processes all signal date locally.