In the heart of Slovenia, researchers at the University of Maribor are revolutionizing the way we think about corn drying, a process crucial for long-term storage and economic viability in the agricultural sector. Marko Simonič, a leading figure from the Faculty of Mechanical Engineering, has spearheaded a groundbreaking study that leverages the power of machine learning to predict moisture content in continuous drying systems. This innovation could significantly impact the energy sector by optimizing drying processes, reducing energy consumption, and enhancing product quality.
Simonič’s research, published in the journal Foods (translated from English), focuses on the application of Long Short-Term Memory (LSTM) neural networks to predict the moisture content of corn at the outlet of continuous drying systems. This approach is part of the broader movement towards Agriculture 4.0, which aims to integrate digital technologies into agricultural practices to meet the growing demand for food efficiently.
The study utilized a dataset of 3826 samples, collected over time from a real-world drying system. Given that the data was not originally intended for predictive modeling, Simonič and his team employed various imputation techniques to ensure data integrity. “The challenge was to make sense of historical data that was not collected with predictive modeling in mind,” Simonič explained. “By applying imputation techniques, we were able to prepare the dataset for efficient modeling with neural networks.”
The LSTM model, which consists of an LSTM layer and three dense layers, was trained on this imputed data. The results were impressive, with the model achieving an RMSE of 0.645, an MSE of 0.416, an MAE of 0.352, and a MAPE of 2.555. These metrics demonstrate high predictive accuracy, indicating that the model can reliably estimate the moisture content of corn based on key process parameters.
The implications of this research are far-reaching. By accurately predicting moisture content, the model can help optimize drying times and temperatures, leading to significant energy savings. “The target air temperature can be minimized, resulting in lower energy consumption,” Simonič noted. “This means fewer natural resources are used, making the process more sustainable.”
Moreover, the model can adjust material discharge intervals to ensure optimal drying conditions, thereby increasing the mass and quality of the product. This not only improves economic profitability but also contributes to the sustainability of the drying process.
Currently, the model is used for offline analysis, allowing operators to evaluate trends in moisture predictions and analyze past drying performance. However, the ultimate goal is to integrate the model into an automated control system for real-time process adjustments. “Future research will focus on developing a fully automated system where the real-time prediction of moisture content dynamically adjusts independent variables such as the target air temperature and material discharge interval,” Simonič said.
The potential for this technology extends beyond corn drying. The principles applied in this study can be adapted to other agricultural processes, paving the way for a more efficient and sustainable future in agriculture. As we move towards Agriculture 4.0, innovations like Simonič’s LSTM model will play a crucial role in optimizing food production and processing, ensuring that we can meet the growing demand for food while minimizing our environmental impact.
