Feature-based classification of representative minerals using laser-induced plasma acoustic signals

Mineral classification is essential in geological exploration and the metallurgical industry. Conventional analytical approaches often require careful sample preparation and complex instrumentation. Laser-induced plasma acoustic (LIPA) signals carry both physical and chemical information about materials, and the acoustic acquisition process requires relatively simple and low-cost instrumentation, making it a potentially useful approach for mineral classification. In this work, the relationships between the LIPA signal features and the spectrum, shock wave velocity, plasma lifetime, plasma radiation temperature, and sample hardness were investigated. LIPA signals were combined with a support vector machine (SVM) algorithm to classify four Fe-rich minerals and three Ca-rich minerals under controlled laboratory conditions. Twenty-six features were extracted from both time-domain and frequency-domain signals. The classification accuracies based on time-domain signals, frequency-domain signals, and the combined feature datasets were 94.29%, 91.43%, and 98.86%, respectively. The classification accuracy was improved by using the feature datasets. Feature selection further indicated that the amplitudes of the time-domain and frequency-domain signals, as well as the delay time of the time-domain signals, are important discriminative features for the minerals investigated in this study. These results suggest that LIPA, combined with SVM algorithm, has potential as a feature-based approach for the classification of representative mineral samples. Further validation on a wider range of minerals and more complex geological samples will be required to assess its broader applicability.