SUMMARY

Computer Numerical Control (CNC) machines are manufacturing equipment that have become indispensable to the modern industrial sphere. CNC machining enables the creation of complex parts with tight tolerances that would have otherwise been unfeasible. Makerspaces are communal workspaces that provide a variety of tools and equipment for use on a wide range of projects. Inclusion of a CNC machine in a makerspace’s lineup of tools supports this goal of variety. However, due to the number of machining parameters that the operator must define for a cut, as well as the specificity of those parameters to the material and project in question, learning to use the CNC presents a steep learning curve for the user. Even with robust training, the user can struggle to choose appropriate settings or can make errors during setup. To reduce this learning curve, users may benefit from feedback on their parameter selection for a cut once it has been run. Previous literature accomplished this by applying several machine learning (ML) models to force data collected from built-in sensors. The performance of the models was evaluated using the area under the receiver operating characteristic curves (AUC). However, many CNC machines lack the built-in sensors utilized in the force-data-based approach and retrofitting one to include such sensors is often prohibitively expensive. The work presented in this thesis investigates the viability of this technique when applied to acoustic data collected from an external sensor. By using an external microphone, the high cost of retrofitting is avoided, making this technology accessible to a wider audience. This investigation demonstrates the feasibility of using an acoustic-data-based ML approach for classifying cuts run on a CNC.