Smart Cyber-Physical System for Pattern Recognition of Illegal 3D Designs in 3D Printing

Abstract

The method to protect intellectual property (IP) in automated manufacturing (AM) and 3D printing industry particularly, presented in this paper, is based on a smart cyber-physical system and the radical improvement of preventive and detective controls to find potential cases of automated manufacturing copyrights infringement. The focus of this paper is not the ecosystem of managing a large network of physical 3D printers, but a smart application and data analysis of data flow within the ecosystem to solve a problem of IP protection and illegal physical objects manufacturing. In this paper, we focus on the first step in this direction – pattern recognition of illegal physical designs in 3D printing, and detection of firearms parts particularly. The proposed method relies on several important steps: normalization of 3D designs, metadata calculation, defining typical illegal designs, pattern matrix creation, new 3D designs challenging, and pattern matrix update. We classify 3D designs into loose groups without strict differentiation, forming a pattern matrix. We use conformity and seriation to calculate the pattern matrix. Then, we perform the analysis of the matrix to find illegal 3D designs. Our method ensures simultaneous pattern discovery at several information levels - from local patterns to global. We performed experiments with 5831 3D designs, extracting 3728 features. It took 12 min to perform pattern matrix calculation based on the test data. Each new 3D design file pattern recognition took 0.32 s on four core, 8 GB ram, 32 GB SSD Azure VM instance.