Issue 78

C. Popa et alii, Fracture and Structural Integrity, 78 (2026) 118-128; DOI: 10.3221/IGF-ESIS.78.07

Figure 4: Heat map of the results for all numerical analysis.

Figure 5: 3D representation of the numerical results highlighting the optimal configuration (red big circle).

M ACHINE L EARNING

Data and features he prediction model was developed using a structured dataset imported from an Excel file and processed within the Python environment. The objective of the dataset preparation stage was to extract the variables that most strongly influence the target shape response. Three independent variables were selected as model inputs: 1) distance A, 2) radius B, 3) angle C. There is a single output variable, representing the shape performance indicator to be predicted. The resulting feature matrix and target vector can be expressed as:   X Input Feature A,Input Feature B,Input Feature C  y Output Feature S  The selected features form the basis for learning the relationship between input parameters and the resulting shape characteristics. Preprocessing and splits Before model training, the dataset underwent preprocessing to prepare the input variables for supervised learning. The extracted features and target values were converted into numerical arrays suitable for machine learning operations. Rows containing incomplete information were considered for exclusion to ensure numerical consistency during model fitting. T

122