AI-Assisted Simulation and Performance Enhancement of Microfluidic Device Systems for Improved Structural Integrity

This study investigates the performance of four machine learning algorithms—Logistic Regression (LR), Support Vector Machine (SVM), Gradient Boosting (GB), and K-Nearest Neighbors (KNN)—for fault detection in an industrial system. The dataset used in this analysis includes synthetic measurements of pressure, flow rate, and temperature, along with binary labels for leakage and deformation, which are indicative of faults. The models were trained on the data after standardizing the features and splitting it into training and test sets. Performance was evaluated using accuracy, area under the curve (AUC), confusion matrices, and ROC curves. Results demonstrate that Gradient Boosting outperforms the other models, achieving the highest accuracy and AUC scores, particularly excelling in handling class imbalances. Logistic Regression and SVM also perform well, with slightly lower AUCs but strong fault detection capabilities. KNN, while effective in some instances, showed weaker performance, especially in terms of AUC and accuracy, likely due to its sensitivity to hyperparameters and feature scaling. This work highlights the efficacy of ensemble methods like Gradient Boosting for fault detection tasks and underscores the importance of model selection and tuning in predictive maintenance applications.