Abstract Coffee is an important agricultural commodity worldwide. Its quality is influenced by several factors that can damage kernels, changing the product composition, flavor, appearance, and causing weight loss. Mathematical models for estimating the dry matter loss of coffee were not found in scientific literature. This study deals with the calibration and validation of a model for predicting the dry matter loss of coffee beans under different storage times, temperatures and moisture contents. For this, green parchment grain of Coffea arabica (cv Catuai Vermelho) were used. Measured dry matter losses varied from 0 to 1.2% and the proposed model was accurate when comparing experimental and predicted data (99.8%). Mean bias, absolute and relative errors were −0.001%, 0.026% and 13.6%, respectively, while the root mean square error was 0.066%, evidencing low scattering and high precision. The estimated maximum allowable storage times of coffee beans, based on a 0.5 dry matter loss threshold, varied from 9 years (for temperature and moisture content of 15 °C and 14% d.b.) to 2 days (for temperature and moisture content of 35 °C and 22% d.b.). The proposed model can be easily implemented in computer programs, appearing as an important tool when simulating the drying or aeration processes, as well as for helping in decision making for coffee storage and trading.

中文翻译：

---

模拟不同储存条件下咖啡豆的干物质损失

摘要 咖啡是世界范围内重要的农产品。其质量受多种因素影响，这些因素会损坏果仁、改变产品成分、风味、外观并导致体重减轻。在科学文献中没有找到用于估计咖啡干物质损失的数学模型。本研究涉及模型的校准和验证，用于预测不同储存时间、温度和水分含量下咖啡豆的干物质损失。为此，使用了小粒咖啡（cv Catuai Vermelho）的绿色羊皮纸颗粒。测量的干物质损失从 0% 到 1.2% 不等，并且在比较实验和预测数据 (99.8%) 时，所提出的模型是准确的。平均偏差、绝对误差和相对误差分别为 -0.001%、0.026% 和 13.6%，而均方根误差为 0.066%，证明低散射和高精度。根据 0.5 干物质损失阈值，估计咖啡豆的最大允许储存时间从 9 年（温度和水分含量为 15 °C 和 14% db）到 2 天（温度和水分含量为 35 °C）不等。 C 和 22% db)。所提出的模型可以很容易地在计算机程序中实现，在模拟干燥或充气过程时，作为一个重要的工具出现，以及帮助制定咖啡存储和交易的决策。