Noncontact ultrasonic materials identification based on improved frequency responses

Compared with existing methods, a possible noncontact ultrasonic materials identification method is proposed by using simple ultrasonic probes off the shelf while keeping a good identification accuracy. In this method, three measures are taken to ensure the good identification effect: 1) Frequency domain filtering including bandpass and notch filtering are used to process the raw ultrasonic echo signals received. The concept of improved frequency responses (IFR) signal is proposed and used as the basis for material identification; 2) Time domain filtering by empirical mode decomposition (EMD) method is adopted for further signal processing. With intrinsic mode function (IMF) obtained, information entropy can be calculated as the final features extracted; 3) By repeatedly testing different samples of the same material, the obtained material's information entropy values are then used to fit a probability density function, which is served as the characteristic template for each material. To demonstrate the feasibility of the proposed method, four aluminum alloy smooth plates of identical size are tested, and an identification accuracy of 100 % is achieved. Due to the low cost of ultrasonic transmitter and receiver, our method can be applied to object identification in many different application scenarios.