Fig. 1 Overview of privacy-preserving collaborative process-defect modeling utilizing ADDAM. This process can be independently applied for multiple AM users, where the de-identified data can be aggregated and used to develop more robust privacy-preserving collaborative models. More

Fig. 2 Current risk of externally shared or aggregated AM process data More

Fig. 3 Key steps of AM and data generated More

Fig. 4 Melt pool angular orientation and printing path: ( a ) θ = 0 deg, ( b ) θ = 60 deg, ( c ) θ = 180 deg, and ( d ) θ = 300 deg More

Fig. 5 Overview of the ADDAM methodology More

Fig. 6 The pseudocode of the ADDAM Algorithm More

Fig. 7 Experimental setup and data collection methods More

Fig. 8 Data splitting for scenario I and scenario II More

Fig. 9 Hyperparameter tuning for distance constraint ( M ) and layer constraint (Δ l ) constraints for scenario I ( a ) and scenario II ( b ) More

Fig. 10 Global k benchmark method tuning for various k values for scenario I ( a ) and scenario II ( b ) More

Fig. 11 Pareto front comparison during the parameter tuning (using T ) for the ADDAM and global k in scenario I (A) and scenario II (B). This details all possible solutions for the different combinations of tuning parameters for ADDAM, and highlights the Pareto optimal solutions, which ... More

Fig. 12 Pareto front comparison evaluating testing data ( E ) using Pareto efficient solutions identified in Fig. 10 , for ADDAM and global k in scenario I (A) and scenario II (B). This details the performance of the optimally selected tuning parameters from Fig. 11 on the testing... More