Modeling Fruit Removal Force and Vibration Thresholds for Selective Harvesting of Robusta Coffee

Abstract

This study developed a preliminary FRF-based model for screening selective vibratory harvesting conditions in robusta coffee. A total of 450 individual fruits were measured from five coffee plants, with 30 red, 30 orange, and 30 green fruits sampled from each plant. For biological-replicate analysis and empirical model fitting, the 30 fruit measurements within each plant-by-maturity combination were averaged, yielding 15 plant-level observations, while the individual-fruit data were retained to characterize within-maturity FRF distributions. Maturity was treated as a categorical predictor with red fruit as the reference class, and fruit mass was included as a covariate. The categorical model closely represented the available plant-level means (R² = 0.9996; RMSE = 0.038 N): after adjustment for fruit mass, orange and green fruits had FRF values 2.433 N and 4.499 N higher than red fruits, respectively, whereas the mass coefficient was small and not significant (p = 0.606). Mean individual-fruit FRF values were 3.533 N for red, 5.966 N for orange, and 8.033 N for green fruits. Using the individual-fruit FRF distributions, the simulation predicted a selective force window of approximately 3.92–5.73 N. Under the idealized point-mass harmonic assumption, this force interval corresponds to approximately 105.4–127.4 Hz at a local amplitude of 3 mm. Because these equivalent frequencies imply very high theoretical local accelerations, they should be interpreted as design-stage screening values rather than field-validated harvester settings.