Developing quality and shelf life predicting models for cherry under map throughout the supply chain

Smit Bipinchandra Patel

doi:10.7273/000007858

Sweet cherries (Prunus avium) are highly perishable fruits whose short shelf life presents significant challenges for postharvest handling and supply chain management. Modified Atmosphere Packaging (MAP) has emerged as a leading technology to extend cherry shelf life by modulating the internal gas composition around the fruit. However, conventional MAP models often fail to account for real-world temperature variability, biological differences such as fruit size, and dynamic changes in package atmosphere, leading to inaccurate shelf-life estimation. This dissertation addresses these limitations by developing and validating predictive models that integrate physiology-based understanding with data-driven machine learning frameworks. The study begins with a comprehensive review of MAP systems, respiration modeling, and shelf-life prediction methods, identifying key gaps in modeling dynamic storage conditions and biological heterogeneity. In Chapter 3, a multi-cultivar analysis introduces a novel size factor to quantify the influence of fruit size on respiration rates, establishing foundational relationships for MAP optimization. Chapter 4 presents predictive models for quality attributes—firmness, hue, and gas composition—developed using deep neural networks (DNN) and one-dimensional convolutional neural networks (1D-CNN), trained on experimental data from Bing cherries stored under MAP. These models estimate quality degradation and residual shelf life with high accuracy under static conditions. In Chapter 5, the study extends to simulate dynamic cold chain environments, including moderate and extreme temperature abuse. Here, a real-time prediction framework is constructed by integrating MAP gas exchange modeling with deep learning architectures (DNN, 1D-CNN, and Generative Adversarial Networks). This approach captures the time–temperature history and gas composition shifts, enabling the estimation of Remaining Shelf Life (RSL) throughout fluctuating conditions. The framework is validated through controlled trials, demonstrating robust performance across variable scenarios. Collectively, this dissertation offers a holistic modeling approach that combines fruit-specific physiology, experimental data, and advanced machine learning to predict quality and shelf life of cherries under MAP across the supply chain. The developed tools have significant implications for reducing food waste, improving logistics, and informing MAP design and cold chain decision-making.

Developing quality and shelf life predicting models for cherry under map throughout the supply chain

Files and links (1)

Abstract

Metrics

Details