A cnn algorithm suitable for the classification of primary and secondary arc-bead and molten mark using laboratory data for cause analysis of electric fires

In this paper, a new CNN algorithm is proposed to determine the direct cause of electric fires. We create 10,000-15,000 three types of data that can occur at a fire scene in our laboratory, and then train and verify it through the proposed CNN algorithm. As a result of the experiment and analysis, the classification accuracy of the primary and secondary arc beads was 86.2%, the accuracy of arc beads and molten marks was 93.6%. And also, the classification accuracy of the primary and secondary arc beads and molten marks was 92.4%. The results of this study are meaningful in that fire forensics can provide accurate identification results in a shorter time through artificial intelligence algorithms compared to the existing methods of identification through visual classification and physicochemical material analysis methods. In particular, the classification between primary and secondary arc beads is known to be a very difficult problem. However, the results of this study provided more than 86% classification ability.