Supply chain optimization for perishable products

Turkey's fresh fruit exports require balancing multiple conflicting objectives such as cost, delivery time, carbon emissions, and freshness due to the perishable nature of the products and the high quality expectations of target markets. In this thesis, an integrated approach has been developed to optimize the sustainable supply chain for citrus and other selected fresh fruits. First, the quality scores of potential orchards were determined using two complementary methods: deep learning–based classification models (MobileNetV2, ResNet50, DenseNet121) for products with sufficient image data, and a fuzzy logic inference system utilizing attributes such as color, surface condition, and shape when data availability was limited. Subsequently, the obtained quality scores were integrated into a multi-objective mathematical model along with capacity, demand, transportation time, cost, and carbon emission data. The model aims to maximize freshness and the quantity of high-quality products while minimizing cost, carbon emissions, delivery time, and the amount of unsold products. Moreover, stochastic deterioration rates were incorporated into the model, and a weight-dependent piecewise cost structure was implemented for air transportation. The model was constructed using real data from a major company exporting fresh fruit from Turkey to 11 markets in Europe and North America. Sensitivity analyses were conducted to examine the impacts of changes in transportation mode preferences, costs, and emission levels. The results demonstrate that quality-oriented solutions provide balanced and sustainable optimal decisions.