ANSYS-Based Simulation and Machine Learning Techniques for Forensic Classification of Knife Wounds

In forensic science, the objective classification of knife-induced injuries is essential for accurately determining the cause and manner of death. Traditional wound analysis often depends on expert judgment, which can be subjective and inconsistent. This paper presents a hybrid framework that combines Finite Element Analysis (FEA) simulations in the Analysis System (ANSYS) with machine learning classifiers in Python to enhance the reliability of forensic wound assessment. Using multilayered tissue models such as skin, fat, and muscle, simulations of stab, incised, and chop wounds were generated under varying knife geometries, insertion angles, and forces. The simulation results, including tissue deformation, stress distribution, and penetration depth, were compiled into a structured dataset. Classifiers such as Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) were evaluated using 5-fold cross-validation. The RF classifier achieved its best performance with 92% accuracy, 0.92 precision, 0.90 recall, and 0.91 F1-score, demonstrating robustness across wound categories. The confusion matrix confirmed high predictive accuracy for stab and incised wounds, with minor misclassifications between deep incised and light chop injuries caused by feature overlap. The proposed system offers forensic experts an interpretable, scalable, and reproducible framework that connects simulation-based biomechanics with data-driven classification. This research emphasizes the potential of combining simulation and machine learning for evidence-based forensic investigations.