Designing mechanical metamaterials to control wave propagation often requires extensive finite element analysis and discrete Fourier transform simulations before fabricating 3D printed structures and conducting experiments. Here, an alternative approach is presented to developing a metamaterial informatics framework by integrating dataset collection with artificial intelligence (AI), which can significantly accelerate the advancement of phononic wave chip technologies based on the triply periodic minimal surface (TPMS). Visualized data analysis is performed to evaluate the sensitivity of phononic band frequency numbers (BNF). Subsequently, various machine learning algorithms are compared for the prediction of sonic BNFs to create a unique identificable encoded mechanical identification tag (EMIT) interacting with sound waves. Then, for the mechanical decoding part with the help of acoustic analogy, a novel concept technology is developed that integrates 3D-printed EMITs with a deep-learning audio classifier for the ownership identification of instruments. Underwater application is discussed further for civil accident investigations, such as echolocating missing aircraft, divers, sunken ships, and containers with valuable cargo. These TPMS-based EMITs represent the first-generation passive sonic frequency identification (SFID) transponder-tags, marking the advent of SFID transponder systems.