In the era of Smart Manufacturing, machining processes are integrated with real-time monitoring systems. This study focuses on precision of surface roughness in the Wire Electrical Discharge Machining (WEDM) process by developing an Artificial Neural Network (ANN) based online surface roughness prediction (OSRP) system. The research begins with the optimization of the machine parameters via Taguchi Design. \(\:{L}_{9}\) Orthogonal array was used in this research allowing for the investigation of 4 controllable factors (Pulse on Time, Pulse off Time, Peak Current and Feed Rate) with 3 different levels each. The optimal parameter settings are identified as 7 µs for Pulse-on Time, 45 µs for Pulse-off Time, 10 A for Peak Current and 0.1 in/min for Feed Rate, resulting in a surface roughness of 78.15 µin reducing approximately 11% this quality feature. A Multi-linear Regression model was developed as a baseline for the article. Then, the Artificial Neural Network model is developed. Both models use as inputs the top three significant output in-process variables from the machine controller which are Maximum Voltage, Current Difference and Maximum Feed Rate derived from regression analysis. The ANN model consisted of a single hidden layer with 128 neurons, trained for 1,700 epochs, Adam optimizer and Mean Absolute Error as loss function. The ANN model’s predictions achieve an average accuracy of 95.8%, as evidenced by comparing with actual measurements. This high accuracy indicates system’s potential to enhance process efficiency by providing real-time predictions, with the in-process variables serving as inputs, ultimately contributing to waste reduction in manufacturing. The system is effective to predict the surface roughness. Furthermore, the ANN-OSRP system allows the operator to monitor the process to accept the piece, reject the piece or fix the parameter settings while the process is taking place.
