Web-Based STEP AP242 Boss and Pocket Feature Recognition for Automated CNC G-Code Generation

This paper presents a web-based island–Pocket Feature Recognition System (IPFRS) that recognises islands and blind pockets from STEP AP242 files and automatically generates CNC milling G-code. The system applies geometric data extraction to parse STEP entities and reconstruct feature boundaries using key geometric definitions such as CARTESIAN_POINT, LINE, CIRCLE, and PLANE. Feature classification is performed by comparing Z-level relationships between reference planes and extracted boundary points to distinguish raised islands (boss) from recessed blind pockets. Based on the recognised features, IPFRS generates toolpaths and converts geometric primitives into standard CNC commands, producing linear and circular motions (e.g., G01, G02/G03) while incorporating user-defined machining parameters such as tool diameter and overlap. The software is implemented using a web architecture (PHP, HTML, and JavaScript) with database support, enabling file upload, processing, code visualisation, and direct G-code download. Validation through CNC simulation using different endmill diameters (5 mm and 10 mm) confirms correct machining sequence, valid syntax, and collision-free toolpaths. The proposed system reduces manual feature identification and programming effort, providing an accessible workflow for consistent G-code generation directly from STEP AP242 models.