Explainable Artificial Intelligence in Computational Materials Science: Mapping the Landscape of CALPHAD-Based Design

Efficient irrigation scheduling is a central challenge in precision agriculture because crop water demand varies with soil properties, climate conditions, growth stage, season, irrigation method, water source, and regional characteristics. This study presents a genetic algorithm-based smart irrigation optimization model that produces operational irrigation recommendations rather than only classifying irrigation need. The model uses an irrigation data set containing 10,000 complete records and 20 variables grouped into soil conditions, weather conditions, agricultural information, and irrigation history. Each candidate irrigation plan is represented as an eight-gene chromosome covering irrigation amount, frequency, time, duration, irrigation type, water source, mulching status, and regional adjustment coefficient. A compound fitness function evaluates candidate plans by combining water adequacy, timing suitability, frequency suitability, duration consistency, method efficiency, source suitability, mulching contribution, regional calibration, soil quality, and excessive water penalty. The evolutionary process uses tournament selection, two-point crossover, adaptive mutation, and elitism. Experimental results show that the population improves across generations and converges toward an interpretable irrigation schedule. The best plan reached a fitness value of 830.2539 in generation 149 and recommended 10.77 mm of water every five days at 18:00 for 14.83 minutes using drip irrigation, rainwater, mulching, and a regional coefficient of 0.91. The results indicate that genetic algorithms can provide a transparent and flexible decision-support structure for smart irrigation planning when multiple agronomic and operational variables must be considered together.