Predicting Corn Moisture Content in Continuous Drying Systems Using LSTM Neural Networks
Abstract
:1. Introduction
2. Materials and Methods
2.1. An Experimental Approach
- The target temperature of the heated air (TA), which determines the level of heat input during the drying process. It is controlled by the burner system, which adjusts the drying temperature to increase or decrease the intensity of moisture removal. A higher target heating air temperature speeds up the drying process but can lead to over-drying or damage to the grains, while a lower temperature can lead to insufficient moisture removal.
- The material discharge interval (DI), which determines how long the corn remains in the drying system before it is discharged to the next stage. It regulates the retention time of the material in the drying modules and, thus, its exposure to heat. A shorter discharge interval increases the material flow but can lead to insufficient drying, while a longer interval enables a longer drying time but can reduce efficiency. The discharge interval is usually set based on the moisture content of the corn in conjunction with the target air temperature—with a higher air temperature allowing a shorter drying time—to achieve optimal drying performance.
- Increased target air temperature (TA): A higher drying temperature accelerates the heating of the module much faster than an adjustment in the material discharge interval (DI). Although the effect is not immediate, TA changes have a faster impact than DI adjustments as the increased heat spreads through the system. In practice, TA is adjusted more frequently, as operators strive to maintain a constant process speed. However, TA is limited to a maximum of 120 °C to prevent grain damage and maintain product quality.
- Extended material discharge interval (DI): While DI adjustments take longer to impact the drying process, they are used when further moisture reduction beyond the TA limit of 120 °C is required. As higher temperatures and rapid drying can have a negative impact on material quality, DI is generally adjusted less frequently than TA.
- Corn level sensor—High: Stops corn transport when the storage reaches maximum capacity (40 tons of corn).
- Corn level sensor—Mid: Triggers the automatic transport of corn into the dryer, which remains active until the material reaches the level sensor—High.
- Level sensor—Low: Stops the drying process due to lack of material and initiates automatic cooling of the dryer.
- Inlet corn moisture (MI);
- Target air temperature (TA);
- Material discharge interval (DI);
- Temperatures of the drying modules (T3, T5, and T6);
- Target variable: Outlet corn moisture (MO).
2.2. Moisture Sampling
2.3. Weather Data Integration
2.4. Preparation of the Datasets
Imputation Methods
- 1.
- Nearest neighbor imputation
- Backward bfill:
- Forward fill:
- Inlet corn moisture (MI): Filled with ffill and bfill, since the corn stored in the silo generally comes from the same batch and has a similar moisture content.
- Module temperatures (T5, T6): Process stability justifies ffill and bfill crediting.
- Material discharge interval (DI): Since the process was mainly regulated by TA, DI was mostly like its nearest neighbors.
- 2.
- Average neighbor imputation
- Target air temperature (TA): Due to frequent process adjustments, a rolling window with 11 data points was used.
- 3.
- Linear interpolation
- Outlet corn moisture (MO): Based on the observed linear relationship.
- 4.
- Regression model
2.5. LSTM Implementation
3. Results and Discussion
- Imputed dataset—missing values were filled using imputation techniques.
- Alternative dataset—rows with missing values were removed.
- Section 3.1 gives a statistical overview of the dataset.
- Section 3.2 analyzes the effects of the weather variables on the performance of the drying system.
- Section 3.3 discusses the results of the model using the imputed dataset and highlights the hyperparameters that perform best.
- Section 3.4 compares the results obtained using the alternative dataset and assesses the impact of handling missing data on model accuracy.
- Section 3.5 outlines the limitations of the study, addressing factors such as data collection methods, environmental variability, and the generalizability of the model.
3.1. Statistical Presentation of the Dataset
- The mean input moisture content (MI) was 25.4%, while the mean output moisture content (MO) was 13.6%, indicating an average moisture reduction of 11.8% during drying. However, the optimal moisture content for storage and processing is between 14% and 14.5%, indicating that the drying process can still be improved to better achieve the target moisture content.
- The discharge interval (DI) averaged 153.2 s, but with a wide range from 90 s to 600 s. The higher values (600 s) only occur during the start-up phase of the dryer when the system is warming up, resulting in temporary fluctuations in drying times before a stable operating state is reached.
- The temperature distribution between the drying modules (T3, T5, T6) shows relatively constant values, although T5 and T6 have higher temperatures than T3. This difference is due to their lower placement in the drying system, where the material remains longer in the dryer and allows for higher heat exposure before reaching the outlet.
3.2. Weather Variables’ Impact on Module Temperatures
- 1.
- Effects of ambient temperature on the drying system:
- An increase in ambient temperature leads to a lower TA (−0.1115). This means that operators can reduce the TA when the ambient temperature is higher.
- An increase in the ambient temperature leads to a higher T3 (+0.1082).
- An increase in the ambient temperature leads to a slightly higher T5 (+0.0148).
- An increase in ambient temperature leads to a lower T6 (−0.1692), which is probably due to the reduction in TA variables.
- Increasing the ambient temperature allows users to reduce the DI (−0.1017).
- 2.
- Effects of relative humidity on the drying system:
- An increase in relative humidity leads to higher TA settings (+0.1131).
- An increase in relative humidity leads to a higher T3 (+0.0576).
- An increase in relative humidity leads to a slightly higher T5 (+0.0090).
- An increase in relative humidity leads to a higher T6 (+0.1329).
- An increase in relative humidity leads to a slight increase in DI (+0.0142).
- An increase in relative humidity leads to a lower MO (−0.0515).
- 3.
- Effects of precipitation on the drying system:
- An increase in precipitation leads to a slight increase in TA (+0.0327), which indicates a small influence on the regulation of air temperature.
- An increase in precipitation leads to a lower T3 (−0.0226), indicating that higher precipitation slightly reduces this process parameter.
- An increase in precipitation leads to a negligible change in T5 (+0.0023), indicating that there is no significant influence.
- An increase in precipitation leads to a slight increase in T6 (+0.0321), indicating that precipitation can slightly influence the heat exchange in the drying process.
- An increase in precipitation leads to a slight increase in DI (+0.0202), indicating that an increase in precipitation may require a slight adjustment in drying intensity.
- An increase in precipitation leads to a slight decrease in MO (−0.0153), indicating that precipitation has little direct influence on the final moisture content.
- 4.
- Effects of solar radiation on the drying system:
- Higher solar radiation leads to lower TA (−0.0932), suggesting that operators can reduce the need for a higher temperature setting due to external heat.
- An increase in solar radiation leads to a lower T3 (−0.0406), indicating minimal direct influence on this process parameter.
- An increase in solar radiation leads to a lower T5 (−0.0072), which indicates little to no direct influence.
- Higher solar radiation leads to a lower T6 (−0.1284), indicating that higher solar radiation and other process adjustments reduce the energy required to maintain this temperature and therefore improve energy efficiency.
- Higher solar radiation leads to a lower DI (−0.0103), indicating a slight reduction in the required drying intensity due to the additional heat input.
- Higher solar radiation leads to a higher MO (+0.0236), indicating that higher solar radiation could slightly increase the final moisture content, although the effect remains minimal.
3.3. Output of Algorithm on Imputed Dataset
3.4. Output of Algorithm on Alternative Dataset
3.5. Study Limitations
- 1.
- Limited moisture measuring points and sampling intervals
- 2.
- Effects of missing data and imputation techniques
- ○
- Temperature values (T3, T5, T6):
- The mean temperature of the drying modules ranged from 49.7 °C (T3) to 64.5 °C (T6), with a standard deviation of 2.15 °C across all modules.
- Missing values in these parameters required imputation, which could result in an error of ±0.2 °C, estimated at 10% of the standard deviation (2.15 °C). Although this error is relatively small in the context of the overall temperature range (20 °C to 71 °C), it could still have a minor effect on the precise control of drying.
- ○
- Moisture content (MI, MO):
- The mean input moisture content (MI) was 25.4%, while the mean outlet moisture content (MO) was 13.6%, corresponding to an average moisture reduction of 11.8%.
- The minimum recorded MO was 10%, while the maximum reached 26%, with the peak occurring during the start-up phase, resulting in transient fluctuations in moisture content prior to stabilization.
- Imputation of the missing moisture values could lead to a shift in model predictions, with potential errors estimated at ±0.3% based on the standard deviation of the recorded moisture values (0.8%). This could affect accuracy, especially near the thresholds for optimal corn storage (14% to 14.5%).
- ○
- Discharge interval (DI):
- DI values range from 90 s to 600 s, with a mean of 153.2 s and a standard deviation of 40.7 s.
- Missing DI values were imputed, resulting in an estimated uncertainty of ±15 s based on the standard deviation of the recorded values.
- This uncertainty in the estimation of dwell time could have an impact on the prediction of drying efficiency, especially when optimizing the discharge timing for uniform moisture removal.
- 3.
- Environmental variability and external conditions
- Ambient temperature (°C): ±2.0 °C variation, affecting the drying air temperature and the moisture evaporation rate.
- Relative humidity (%): ±5.0% deviation, which affects the equilibrium moisture content of the corn and the overall drying efficiency.
- Precipitation (mm): ±1.5 mm variation, which can change the humidity conditions in the environment and affect the characteristics of the drying air.
- Solar radiation (W/m²): ±20 W/m2 variation, resulting in fluctuations in external heat input that can affect the required drying intensity.
- These discrepancies lead to potential variations in drying efficiency estimates and affect the prediction accuracy of the model in real applications. Future improvements should consider on-site environmental monitoring to minimize external measurement discrepancies.
- 4.
- Generalization and applicability to other drying systems
- 5.
- The impact of the frequency of data collection on the performance of the model
- 6.
- The influence of the accuracy of the measuring devices
4. Conclusions
- The experimental approach of recording process parameters in addition to moisture measurements is critical to accurately predicting the initial moisture content of corn in continuous drying systems.
- By applying imputation techniques, it is possible to prepare a human-dependent dataset for efficient modeling with neural networks.
- LSTM neural networks have proven to be a highly effective method for predicting moisture content based on key variables, including the target air temperature (TA), material discharge interval (DI), inlet moisture content, drying module temperatures (T3, T5, T6), and environmental conditions.
- The performance of the model was evaluated using various statistical metrics: mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The results indicate excellent predictive accuracy, as evidenced by the following metrics: RMSE = 0.645, MSE = 0.416, MAE = 0.352, and MAPE = 2.555.
- The target air temperature can be minimized, resulting in lower energy consumption, so that fewer natural resources are used.
- The material discharge intervals can be adjusted as required so that drying times are set to the optimal conditions and more of the product can be processed.
- The desired moisture content at the outlet of the drying system results in a greater mass and higher quality of the product.
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Frank, A.G.; Dalenogare, L.S.; Ayala, N.F. Industry 4.0 Technologies: Implementation Patterns in Manufacturing Companies. Int. J. Prod. Econ. 2019, 210, 15–26. [Google Scholar] [CrossRef]
- Barreto, L.; Amaral, A.; Pereira, T. Industry 4.0 Implications in Logistics: An Overview. Procedia Manuf. 2017, 13, 1245–1252. [Google Scholar] [CrossRef]
- Carou, D. Aerospace Transformation through Industry 4.0 Technologies. In Aerospace and Digitalization: A Transformation Through Key Industry 4.0 Technologies; Springer International Publishing: Cham, Switzerland, 2021; pp. 17–46. [Google Scholar]
- Ghobakhloo, M.; Fathi, M. Industry 4.0 and Opportunities for Energy Sustainability. J. Clean. Prod. 2021, 295, 126427. [Google Scholar] [CrossRef]
- Arden, N.S.; Fisher, A.C.; Tyner, K.; Yu, L.X.; Lee, S.L.; Kopcha, M. Industry 4.0 for Pharmaceutical Manufacturing: Preparing for the Smart Factories of the Future. Int. J. Pharm. 2021, 602, 120554. [Google Scholar] [CrossRef]
- Kamber, E.; Sonmezturk Bolatan, G.I. Industry 4.0 Concept and Applications on Different Sectors. J. Glob. Strateg. Manag. 2020, 14, 31–44. [Google Scholar] [CrossRef]
- Liu, Y.; Ma, X.; Shu, L.; Hancke, G.P.; Abu-Mahfouz, A.M. From Industry 4.0 to Agriculture 4.0: Current Status, Enabling Technologies, and Research Challenges. IEEE Trans. Ind. Inform. 2021, 17, 4322–4334. [Google Scholar] [CrossRef]
- Albiero, D.; Paulo, R.L.D.; Félix Junior, J.C.; Santos, J.D.S.G.; Melo, R.P. Agriculture 4.0: A Terminological Introduction. Rev. Ciênc. Agron. 2020, 51, e20207737. [Google Scholar] [CrossRef]
- Kovács, I.; Husti, I. The Role of Digitalization in Agriculture 4.0—How to Connect Industry 4.0 to Agriculture? Hung. Agric. Eng. 2018, 33, 38–42. [Google Scholar] [CrossRef]
- Davis, K.F.; Gephart, J.A.; Emery, K.A.; Leach, A.M.; Galloway, J.N.; D’Odorico, P. Meeting Future Food Demand with Current Agricultural Resources. Glob. Environ. Change 2016, 39, 125–132. [Google Scholar] [CrossRef]
- Darrah, L.L.; McMullen, M.D.; Zuber, M.S. Breeding, genetics, and seed corn production. In Corn, 3rd ed.; AACC International Press: St. Paul, MN, USA, 2019; pp. 19–41. [Google Scholar]
- Jiang, H.; Hu, H.; Zhong, R.; Xu, J.; Xu, J.; Huang, J.; Wang, S.; Ying, Y.; Lin, T. A deep learning approach to conflating heterogeneous geospatial data for corn yield estimation: A case study of the US Corn Belt at the county level. Glob. Change Biol. 2019, 26, 1754–1766. [Google Scholar] [CrossRef]
- Yang, W.; Nigon, T.; Hao, Z.; Paiao, G.D.; Fernández, F.G.; Mulla, D.; Yang, C. Estimation of corn yield based on hyperspectral imagery and convolutional neural network. Comput. Electron. Agric. 2021, 184, 106092. [Google Scholar] [CrossRef]
- Zhang, R.; Ma, S.; Li, L.; Zhang, M.; Tian, S.; Wang, D.; Liu, K.; Liu, H.; Zhu, W.; Wang, X. Comprehensive utilization of corn starch processing by-products: A review. Grain Oil Sci. Technol. 2021, 4, 89–107. [Google Scholar] [CrossRef]
- Wei, S.; Xie, W.; Zheng, Z.; Yang, D. Numerical and experimental studies on drying behavior of radio frequency-assisted convective drying for thin-layer corn kernels. Comput. Electron. Agric. 2021, 191, 106520. [Google Scholar] [CrossRef]
- Ren, G.Y.; Zhang, L.D.; Zeng, F.L.; Li, Y.B.; Li, L.; Duan, X. Effects of hot air drying temperature and tempering time on the properties of maize starch. Int. J. Agric. Biol. Eng. 2020, 13, 236–241. [Google Scholar] [CrossRef]
- Rogovskii, I.L.; Titova, L.L.; Trokhaniak, V.I.; Solomka, O.V.; Popyk, P.S.; Shvidia, V.O.; Stepanenko, S.P. Experimental studies on drying conditions of grain crops with high moisture content in low-pressure environment. INMATEH-Agric. Eng. 2019, 57, 141–146. [Google Scholar] [CrossRef] [PubMed]
- Liu, Z.; Xu, Y.; Han, F.; Zhang, Y.; Wang, G.; Wu, Z.; Wu, W. Control method for continuous grain drying based on equivalent accumulated temperature mechanism and artificial intelligence. Foods 2022, 11, 834. [Google Scholar] [CrossRef]
- Abdoli, B.; Zare, D.; Jafari, A.; Chen, G. Evaluation of the air-borne ultrasound on fluidized bed drying of shelled corn: Effectiveness, grain quality, and energy consumption. Dry. Technol. 2018, 36, 1749–1766. [Google Scholar] [CrossRef]
- Omari, A.; Behroozi-Khazaei, N.; Sharifian, F. Drying kinetic and artificial neural network modeling of mushroom drying process in microwave-hot air dryer. J. Food Process Eng. 2018, 41, e12849. [Google Scholar] [CrossRef]
- Kalathingal, M.S.H.; Basak, S.; Mitra, J. Artificial neural network modeling and genetic algorithm optimization of process parameters in fluidized bed drying of green tea leaves. J. Food Process Eng. 2020, 43, e13128. [Google Scholar] [CrossRef]
- Liu, Z.L.; Bai, J.W.; Wang, S.X.; Meng, J.S.; Wang, H.; Yu, X.L.; Xiao, H.W. Prediction of energy and exergy of mushroom slices drying in hot air impingement dryer by artificial neural network. Dry. Technol. 2020, 38, 1959–1970. [Google Scholar] [CrossRef]
- Azadbakht, M.; Aghili, H.; Ziaratban, A.; Vahedi Torshizi, M. Application of artificial neural network method to exergy and energy analyses of fluidized bed dryer for potato cubes. Energy 2017, 120, 947–958. [Google Scholar] [CrossRef]
- Smys, S.; Chen, J.I.Z.; Shakya, S. Survey on neural network architectures with deep learning. J. Soft Comput. Paradig. 2020, 2, 186–194. [Google Scholar]
- Jin, Y.; Wong, K.W.; Yang, D.; Zhang, Z.; Wu, W.; Yin, J. A neural network model used in continuous grain dryer control system. Dry. Technol. 2022, 40, 1901–1922. [Google Scholar] [CrossRef]
- Chen, C.; Zhang, W.; Shan, Z.; Zhang, C.; Dong, T.; Feng, Z.; Wang, C. Moisture contents and product quality prediction of Pu-erh tea in sun-drying process with image information and environmental parameters. Food Sci. Nutr. 2022, 10, 1021–1038. [Google Scholar] [CrossRef]
- Noh, S.-H. Analysis of gradient vanishing of RNNs and performance comparison. Information 2021, 12, 442. [Google Scholar] [CrossRef]
- Zhou, H.; Huang, M.; Zhu, Q.; Zhang, M. Developing C-LSTM model for evaluating moisture content of carrot slices during drying. Dry. Technol. 2022, 40, 2964–2972. [Google Scholar] [CrossRef]
- Zhang, Y.; Fang, Z.; Li, C.; Li, C. Deep-Learning-Based Model Predictive Control of an Industrial-Scale Multistate Counter-Flow Paddy Drying Process. Foods 2024, 13, 43. [Google Scholar] [CrossRef]
- Meng, J.; Li, C.; Tao, J.; Li, Y.; Tong, Y.; Wang, Y.; Zhang, L.; Dong, Y.; Du, J. RNN-LSTM-based model predictive control for a corn-to-sugar process. Processes 2023, 11, 1080. [Google Scholar] [CrossRef]
- Sabat, M.; Kotwaliwale, N.; Chakraborty, S.K.; Kumar, M. Long short-term memory-based real-time monitoring of potato slice drying using image chromatic features. J. Food Process. Preserv. 2022, 46, e17232. [Google Scholar] [CrossRef]
- FOSS Analytical Solutions for Food Quality Improvement and Control. Available online: https://www.fossanalytics.com (accessed on 4 April 2024).
- Bureau Veritas Slovenia. Available online: https://www.bureauveritas.si/ (accessed on 4 April 2024).
- Chen, J.; Shao, J. Nearest neighbor imputation for survey data. J. Off. Stat. 2000, 16, 113. [Google Scholar]
- Noor, N.M.; Al Bakri Abdullah, M.M.; Yahaya, A.S.; Ramli, N.A. Comparison of linear interpolation method and mean method to replace the missing values in environmental data set. Mater. Sci. Forum 2015, 803, 278–281. [Google Scholar] [CrossRef]
- Alita, D.; Putra, A.D.; Darwis, D. Analysis of classic assumption test and multiple linear regression coefficient test for employee structural office recommendation. Indones. J. Comput. Cybern. Syst. 2021, 15, 295. [Google Scholar] [CrossRef]
- Simonič, M. Analysis and Optimization of Process Parameters in a Corn Drying System: Master’s Thesis. Master’s Thesis, University of Maribor, Maribor, Slovenia, 2024. Available online: https://dk.um.si/IzpisGradiva.php?lang=slv&id=87358 (accessed on 18 February 2025).
- Xing, S.; Lin, Z.; Gao, X.; Wang, D.; Liu, G.; Cao, Y.; Liu, Y. Research on Outgoing Moisture Content Prediction Models of Corn Drying Process Based on Sensitive Variables. Appl. Sci. 2024, 14, 5680. [Google Scholar] [CrossRef]
- Li, B.; Li, C.; Huang, J.; Li, C. Application of Artificial Neural Network for Prediction of Key Indexes of Corn Industrial Drying by Considering the Ambient Conditions. Appl. Sci. 2020, 10, 5659. [Google Scholar] [CrossRef]
Parameter | Value/Range | Description |
---|---|---|
Drying process | Continuous Convective | Downward material flow through drying modules |
Material flow rate | 9–12 t/h | Controlled by discharge interval |
Discharge interval (DI) | Adjustable (120–160 s, avg: 153.2 s) | Regulates corn retention time |
Target air temperature (TA) | Adjustable (Max: 120 °C) | Higher TA speeds up drying but is limited to avoid grain damage |
Module temperatures (mean) | T3: 49.7 °C, T5: 59.5 °C, T6: 64.5 °C | Heat distribution across drying tower |
Inlet moisture (MI) | 25.4% (mean) | Mean moisture value before drying |
Outlet moisture (MO) | 13.6% (mean) | Mean moisture value after drying |
Column | MI | TA | T3 | T5 | T6 | DI | MO |
---|---|---|---|---|---|---|---|
Missing data (%) | 5.18 | 0.78 | 28.94 | 1.07 | 1.12 | 0.97 | 1.86 |
Label | Symbol in Equation | Coefficient/Value |
Intercept () | β0 | 29.65 |
Average air temperature—β1 | Tavg (β1) | 0.14 |
Average relative humidity (%) | RHavg (β2) | 0.03 |
Precipitation (mm) | P (β3) | −0.43 |
Global radiation (W/m2) | Radglob (β4) | −0.01 |
Inlet moisture | MI (β5) | −0.02 |
Target air temperature | TA (β6) | 0.045 |
Module 5 temperature | T5 (β7) | 0.01 |
Module 6 temperature | T6 (β8) | 0.174 |
Material discharge interval | MD (β9) | −0.002 |
Optimizers | Adam, Adagrad, Ftrl, Adadelta, Nadam |
Activations | Relu, Prelu, Elu, Selu, Swish |
Learning rates | 0.01, 0.001 |
Epochs | 16, 64, 100 |
Batch sizes | None, 16, 32, 64 |
MI | TA | T3 | T5 | T6 | DI | MO | |
---|---|---|---|---|---|---|---|
Mean | 25.4 | 111.7 | 49.7 | 59.5 | 64.5 | 153.2 | 13.6 |
Std Dev | 2.5 | 7.4 | 2.15 | 3.61 | 4.1 | 40.7 | 0.8 |
Min | 14.0 | 60.0 | 20.0 | 19.0 | 18.0 | 90.0 | 10.0 |
25th Percentile | 23.3 | 110 | 48.9 | 58.0 | 63.0 | 135.0 | 13.2 |
50th percentile | 25.3 | 110 | 49.8 | 60.0 | 65.0 | 145.0 | 13.6 |
75th Percentile | 26.7 | 118 | 50.6 | 61.0 | 67.0 | 160.0 | 13.9 |
Max | 36.8 | 125 | 71.0 | 82.0 | 85.0 | 600.0 | 26.0 |
Activation | Optimizer | Learning Rate | Epochs | Batch Size | MSE | RMSE | MAE | MAPE |
---|---|---|---|---|---|---|---|---|
PReLU | Adam | 0.001 | 16 | 32 | 0.418 | 0.647 | 0.352 | 2.554 |
elu | Adam | 0.001 | 16 | 16 | 0.418 | 0.647 | 0.352 | 2.553 |
elu | Adagrad | 0.001 | 100 | 64 | 0.419 | 0.647 | 0.352 | 2.555 |
elu | Adadelta | 0.001 | 64 | 32 | 0.419 | 0.647 | 0.352 | 2.555 |
elu | Adadelta | 0.001 | 64 | 64 | 0.418 | 0.647 | 0.352 | 2.557 |
elu | Nadam | 0.001 | 16 | 16 | 0.418 | 0.647 | 0.352 | 2.554 |
elu | Ftrl | 0.001 | 64 | 16 | 0.419 | 0.647 | 0.352 | 2.558 |
swish | Ftrl | 0.01 | 16 | 32 | 0.417 | 0.646 | 0.352 | 2.552 |
swish | Adam | 0.001 | 16 | None | 0.416 | 0.645 | 0.352 | 2.555 |
swish | Nadam | 0.001 | 16 | None | 0.419 | 0.647 | 0.352 | 2.553 |
Activation | Optimizer | Learning Rate | Epochs | Batch Size | MSE | RMSE | MAE | MAPE |
---|---|---|---|---|---|---|---|---|
Relu | Adam | 001 | 16 | 32 | 12.38 | 3.52 | 0.428 | 2.49 |
Relu | Adam | 0.01 | 64 | 64 | 12.37 | 3.52 | 0.432 | 2.53 |
Relu | Nadam | 0.01 | 16 | 64 | 12.46 | 3.53 | 0.432 | 2.51 |
Relu | Nadam | 0.01 | 64 | 64 | 12.37 | 3.52 | 0.428 | 2.49 |
PReLU | Ftrl | 0.01 | 64 | 32 | 12.35 | 3.51 | 0.439 | 2.58 |
PReLU | Ftrl | 0.01 | 100 | 16 | 12.34 | 3.51 | 0.439 | 2.58 |
Elu | Nadam | 0.001 | 64 | None | 12.45 | 3.53 | 0.433 | 2.52 |
Selu | Adagard | 0.01 | 16 | 16 | 12.48 | 3.53 | 0.429 | 2.46 |
Selu | Ftrl | 0.01 | 16 | 16 | 12.5 | 3.54 | 0.436 | 2.54 |
Selu | Nadam | 0.001 | 100 | 16 | 12.45 | 3.53 | 0.43 | 2.5 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Simonič, M.; Ficko, M.; Klančnik, S. Predicting Corn Moisture Content in Continuous Drying Systems Using LSTM Neural Networks. Foods 2025, 14, 1051. https://doi.org/10.3390/foods14061051
Simonič M, Ficko M, Klančnik S. Predicting Corn Moisture Content in Continuous Drying Systems Using LSTM Neural Networks. Foods. 2025; 14(6):1051. https://doi.org/10.3390/foods14061051
Chicago/Turabian StyleSimonič, Marko, Mirko Ficko, and Simon Klančnik. 2025. "Predicting Corn Moisture Content in Continuous Drying Systems Using LSTM Neural Networks" Foods 14, no. 6: 1051. https://doi.org/10.3390/foods14061051
APA StyleSimonič, M., Ficko, M., & Klančnik, S. (2025). Predicting Corn Moisture Content in Continuous Drying Systems Using LSTM Neural Networks. Foods, 14(6), 1051. https://doi.org/10.3390/foods14061051