Abstract
Heat treatment during storage is effective in delaying the ripening of fruit. In this study, an optimal pattern of heat treatment for tomatoes was investigated using an intelligent control technique consisting of neural network and genetic algorithm. An objective function was given by the reciprocal number of the color development from green to red for evaluating the ripening of tomatoes. The control process was divided into 8 steps. First, the color development was identified using neural network. Then, the set point of temperature which maximized the objective function at each of the 8 steps was sought through simulation of the identified model using genetic algorithm. The genetic algorithm allowed an optimal heat treatment to be successfully sought through simulation of an identified neural network model. The optimal heat treatment obtained here was intermittent, while the normal heat treatment was only one treatment at the first stage. Finally this optimal heat treatment was applied to an actual system. The results show that the optimal heat treatment indicates a better result in maintaining the color development of fruit than the conventional one.
Introduction
Post-harvest ripening management is critical for extending the shelf life of produce. For tomatoes (Lycopersicon esculentum Mill.), the transition from mature green to red involves complex biological processes, including ethylene production and lycopene synthesis. While refrigeration is the standard method for slowing these processes, heat treatment has emerged as a promising alternative. Short bursts of heat can inhibit ripening enzymes and trigger the production of protective heat shock proteins (HSPs).
However, the biological response of fruit to thermal stress is non-linear and highly complex, making it difficult to determine an optimal temperature schedule through simple trial-and-error. This study proposes an intelligent control approach to discover a temperature pattern that maximizes the duration tomatoes remain in the "mature green" stage.
Methods: The Speaking Fruit Approach
The researchers utilized the "Speaking Fruit Approach" (SFA), which treats the fruit as an active participant in the control loop. The storage environment is dynamically adjusted based on the fruit's predicted physiological response.
The computational strategy involved two primary stages:
- System Identification (Neural Networks): A three-layer neural network modeled the tomato’s color development in response to temperature. By training on historical data (past temperatures and color states), the model learned to predict future ripening. Cross-validation determined that a model with 4 hidden neurons provided the minimum error.
- Optimization (Genetic Algorithms): Using the neural network as a digital simulator, a Genetic Algorithm (GA) "evolved" the optimal 8-day temperature schedule. The GA evaluated thousands of 8-step sequences, using the reciprocal of color development as an objective function to favor schedules that minimized ripening.
Results and Discussion
The most significant finding was that the optimal schedule was not a single initial heat shock, but an intermittent pattern of periodic heat spikes.
In simulations, this intermittent heat treatment (periodic spikes to 35°C) significantly outperformed both constant cooling (5°C) and the conventional "normal" heat treatment (a single 35°C burst followed by 5°C). The researchers hypothesize that these periodic spikes repeatedly trigger the synthesis of HSPs, helping the fruit resist ripening stress more effectively than a single treatment.
When applied to actual tomatoes in a storage chamber, the optimized intermittent treatment successfully maintained fruit "greenness" better than conventional cooling. While the real-world difference between intermittent and single-burst treatments was less pronounced than in simulation, the intermittent approach showed a more consistent ability to suppress ripening over the full 8-day period.
Conclusion
This study demonstrates that combining neural networks and genetic algorithms can successfully identify non-obvious, optimal environmental controls for biological systems. The discovery of the effectiveness of intermittent heat treatment provides a new strategy for post-harvest storage. Future work will explore higher temperature ceilings and the scalability of this intelligent control system for industrial-scale warehouse management.
Summary updated