Application of D-Optimal Mixture Design and GA-ANFIS–PSO Hybrid Models for Predicting and Optimizing Sensory Properties and Nitrite Residues in Probiotic Fermented Sausage Production

This study optimized the sensory properties and nitrite residues in probiotic fermented sausages using a D-Optimal Mixture Design enhanced by GA-ANFIS (Genetic Algorithm-Adaptive Neuro-Fuzzy Inference System) and PSO-ANFIS (Particle Swarm Optimization Adaptive Neuro-Fuzzy Inference System) hybrid models. The optimization criteria were to minimize nitrite residue while maximizing sensory properties. In the D-Optimal Mixture Design, the following the three constraints used were Lallemand – CXSP (X1) with a range of 0% to 0.04%, Bitec LS 25 (X2) with a range of 0% to 0.04% and L. Acidophilus Nu-TrishLA-5 (X3) with a range of 0% to 0.04%. Additionally, the total components must sum to 0.04%, represented as (X1 + X2 + X3 = 0.04%). Sixteen formulations were tested for their color, flavor, chewiness, and overall taste on a 5-point scale, while also measuring nitrite levels. The best formulation contained either 0.04% L. acidophilus Nu-TrishLA-5 or a mix of 0.024% L. acidophilus Nu-TrishLA-5 and 0.016% Lallemand – CXSP, achieving lower nitrite levels (12.73 to 19 mg/kg) than regular non-fermented Iranian sausages. The GA-ANFIS and PSO-ANFIS models showed better accuracy, with higher coefficient of determination (R2) and lower Root Mean Square Error (RMSE) values than the D-Optimal models, but all models provided reliable results. This approach offers a way to formulate fermented sausages that reduce nitrite content while maintaining desirable sensory characteristics.