Smart Sensor-Based Monitoring System for Fresh Produce Shelf Life Prediction

Thesis Abstract

The rapid deterioration of fresh produce during transportation and storage presents significant challenges to supply chain efficiency, food safety, and consumer satisfaction, necessitating innovative solutions for real-time quality monitoring. This study aims to develop a smart sensor-based monitoring system capable of accurately predicting the remaining shelf life of perishable produce such as strawberries, tomatoes, and leafy greens. The specific objectives include identifying critical spoilage indicators through sensor data, designing and integrating a prototype sensor network, and employing advanced data analytics to model shelf life prediction. The research adopts a mixed-methods approach, combining quantitative sensor data collection with qualitative assessments to validate the system’s accuracy and usability. The research design primarily involves a longitudinal experimental protocol conducted at a commercial produce storage facility, involving a sample size of 300 produce items, equally distributed across three categories—strawberries, tomatoes, and leafy greens. Sensors measuring temperature, humidity, volatile organic compounds (VOCs), ethylene levels, and pH are attached to representative samples, with data collected at 30-minute intervals over a period of 14 days. To facilitate comprehensive analysis, a combination of statistical techniques such as multiple linear regression, principal component analysis (PCA), and support vector machine (SVM) modeling is employed to identify significant spoilage predictors and develop an optimized predictive algorithm. The sensor data are complemented by microbiological assessments and visual quality scoring performed bi-daily to establish ground truth for spoilage status. Key findings are anticipated to demonstrate strong correlations between specific sensor measurements—particularly VOC concentrations, ethylene emissions, and temperature fluctuations—and actual spoilage progression, thereby enabling the creation of a robust predictive model for shelf life estimation. The predictive accuracy of the system is expected to achieve an R-squared value exceeding 0.85, with precision metrics indicating high reliability in predicting remaining freshness under diverse storage conditions. The results are further analyzed using receiver operating characteristic (ROC) curves to assess the model’s sensitivity and specificity, ensuring practical applicability in real-world scenarios. This research contributes novel insights into integrating sensor technologies with machine learning techniques specifically tailored for perishable food monitoring, filling notable gaps in current literature which predominantly focus on post-harvest assessment rather than real-time predictive systems. The study advances the theoretical understanding of spoilage dynamics, supported by the application of the Theory of Diffusion of Innovation to facilitate the adoption of sensor-based solutions within supply chains. A conceptual model illustrating the relationship between sensor signals, environmental conditions, and spoilage indicators is proposed, serving as a foundation for future developments in intelligent food logistics. The study concludes that a sensor network combined with predictive analytics provides a feasible, non-destructive, and timely method for estimating produce shelf life, thereby enabling better inventory management and reducing food waste. Recommendations include scaling the prototype for commercial deployment, integrating wireless data transmission modules for remote monitoring, and developing user-friendly interfaces for stakeholders along the supply chain. The study also suggests avenues for further research, such as expanding the range of sensors, exploring other perishables, and applying deep learning algorithms for more complex predictive modeling. Overall, this work demonstrates the potential of ICT-driven approaches to enhance the sustainability and safety of food distribution systems effectively.