Parameter Estimation for Ultrasonics Echoes Using an Weighted Mean of Vectors Optimizer

Accurate estimations of the parameters of the ultrasonic echo pattern are essential in ultrasonic nondestructive testing. The estimation of this parameters allow characterization and defect detection in the materials. However, estimations the parameters of multi-echo ultrasonic signals is a challenging task in the cases of closely spaced echoes and/or drowned in noise. Therefore, this paper proposes a potent integrated algorithm for estimating parameters of multi-echo ultrasonic signals using an optimizer called "weighted mean of vectors" (INFO) and the principle of minimum description length (MDL). The INFO algorithm is an optimizer that uses the concept of weighted average to move agents to a better position. It modified the weighted average method by using three central processes, namely the update rule, vector combination, and the local search. The principle of MDL is used to determine the number of echoes, i.e., the order of the model. A simulation study has been carried out simulating a signal containing three echoes that overlap in time with several levels of noise. Additionally, experimental tests were performed on three steel samples, each containing two adjacent holes drilled in the back wall face. Both experimental and simulated results show that the proposed method can accurately estimate the parameters of closely spaced echoes.