Neural network modeling of process parameters for electrical discharge machining

A neural network model was developed to predict the quality attributes for electrical discharge machining (EDM). During this study, in-situ monitoring of several transient parameters was used to describe the process, which includes current, displacement, and voltage. The aforementioned parameters were translated using linear regression and utilized as the input to the neural network. The quality attributes of the process, break-through and airflow measurement were used as the model's target. This concept is illustrated in Figure 1. Break-through is the term given to the condition when the airflow cavities in the product have been completely eroded through. The airflow is a measure that defines the mass flow rate through these cavities manufactured using EDM. Due to the unreliable nature of the EDM process to create the desired quality in the product, post inspection is required for both break-through and airflow. This involves manually probing each hole with a pin to ensure full break-through. If every one of the approximately 65 holes has complete break-through, then the part needs to be prepared for the airflow inspection at a separate workbench. This includes washing the part and applying wax to certain holes. Then the part is tested on the air flow bench and subsequently heated to remove the wax. The post inspection is very costly; requiring time, labor, material, and manufacturing floor space. The neural network model, is capable of predicting the quality attributes. As a result, this eliminates the need for post inspection.; Since the quality attributes can be determined in real-time, future research will entail evolving the model into an intelligent control system. Implemented in a control system, this research will be used to ensure the quality attributes manufactured into the product. That is, it will prevent defects from occurring as a result of EDM process randomness.