Nowadays, there is a general understanding that artificial intelligence (AI) has great potential to improve drying efficiency and product quality. Since intelligence is usually understood as the ability to learn from the environment, one of the central topics of AI is machine learning (ML).[Citation1] However, the application of AI for the control of drying is not limited by ML algorithms. Intelligent drying requires also setting of specific objectives, e.g., energy efficiency, product quality, or both.[Citation2] Objective-oriented optimization needs a specific drying strategy. This aspect of AI implementation is typically missing in most reviews on the topic. This Editorial will focus on drying strategies as the critical step in the development of intelligent drying systems.
It is important to trace the evolution of drying strategies in a historical context. The first “primitive” drying pursued only one objective – to decrease the water content to a safe level for long-term storage. Open-air natural drying, using wind and sun, served well for this objective and satisfied people in the pre-industrial era. Since drying occurred under ambient conditions, it did not require any control.
The era of industrial drying can be traced back to the industrial revolution of the late eighteenth century. The need to dry materials such as textiles, grains and chemicals, on a large scale led to the development of industrial drying methods. By the nineteenth century, various drying techniques, including the use of steam, hot air and specialized drying equipment, became integral to industrial processes. The objective of industrial drying significantly changed to increase the drying rate through additional convective, conductive or radiative heat supply. For this purpose, a simple drying strategy with a constant heat supply was adopted. A corresponding control strategy required keeping the constant elevated temperature of the environment. This strategy assumed that the temperature of the material was constant during the entire period of drying, which would guarantee constant heat transfer. However, this was not the case. The transient nature of drying was recognized by early researchers in the field of drying.[Citation3,Citation4] Moreover, unpredictable changes in thermophysical properties posed significant challenges for the control of drying. Over the years, researchers and engineers have employed various strategies to address this complexity and improve drying efficiency. One example is intermittent drying, applied for heat-sensitive biomaterials with varying drying characteristics.[Citation5] Another example is a stepwise control strategy, addressing the changes in material thermophysical properties during drying.[Citation6] This strategy was successfully applied for ginseng root and carrot drying, improving drying efficiency and product quality.[Citation7,Citation8] Transient behavior reveals itself as an increase in material temperature in the falling rate period of drying. Hence, the strategy of constant heat supply led to increased heat losses and possible overheating of a material. As an alternative, another control strategy of variable heat supply was proposed.[Citation9] This control strategy was based on the continuous observation of material temperature and synchronizing heat transfer with mass transfer. This control strategy was beneficial due to higher efficiency of energy use and less risk of quality damage.
Increased requirements on process efficiency and product quality pose new challenges for the drying industry and demonstrate a growing need for optimization of control strategies.[Citation10] Multi-objective optimization requires advanced control strategies that adapt to the changing conditions during drying. For example, a multi-objective optimization, minimizing energy consumption while maximizing vitamin content, was proposed.[Citation11]
The progress in AI technologies allows the combination of multiple AI tools to implement the so-called adaptive control. For example, the combination of ANN, GA and Pareto strategy was successfully used for apples and blueberry drying.[Citation10,Citation11] Another example is the combination of ANN, GA and FL for unsupervised adaptive control of drying conditions to maximize product quality.[Citation12] Adaptive control involves a feedback loop that adjusts drying parameters based on real-time measurements of product quality.
Both multi-objective optimization and adaptive control represent intelligent control strategies, based on the real-time monitoring of product quality. If monitoring is accomplished with elements of AI, such as soft sensors and ML algorithms, it could be called an “intelligent observer.” At the initial stage of development, intelligent observer could act as a digital researcher, exploring the cause-effect links between drying conditions and product quality. Big datasets, collected during the operation, could be used for fine-tuning drying strategies for a specific product. The implementation of intelligent observer alone will benefit the drying industry because of significant savings in the research.
The second stage could be the utilization of intelligent observer for the optimization of the drying regime. At first, it can be done in the “advisory” mode, to help an operator to develop the best drying strategy and recognize the benefit of dynamic optimization for saving time, energy or product quality. After control strategies are well understood, the intelligent observer will become part of the adaptive control system. For example, the control strategy matching diffusive and convective mass transfer will include simultaneous control of temperature, humidity and airflow. Setting limits and alarm modes will reduce the load on the operator and minimize risks of product damage.
It has been shown that the integration of machine vision and machine learning with an AI decision-making framework has significant potential to improve the efficiency of industrial drying systems.[Citation12] The importance of adaptive control strategies is increasing because of the need for more sustainable and energy-efficient drying.
Dalhousie University, Canada
[email protected]
Arun Mujumdar
McGill University, Canada
Disclosure statement
No potential conflict of interest was reported by the author(s).
References
- Martynenko, A.; Misra, N. N. Machine Learning in Drying. Drying Technol. 2020, 38, 596–609. DOI: 10.1080/07373937.2019.1690502.
- Intelligent Control in Drying (Ed. by A. Martynenko, A. Buck). CRC Press, Boca Raton, 2019.
- Lewis, W. K. The Rate of Drying of Solid Materials. J. Ind. Eng. Chem. 1921, 13, 427–432. DOI: 10.1021/ie50137a021.
- Sherwood, T. K. The Drying of Solids—II. Ind. Eng. Chem. 1929, 21, 976–980. DOI: 10.1021/ie50238a021.
- Chua, K. J.; Mujumdar, A. S.; Chou, S. K. Intermittent Drying of Bioproducts – An Overview. Bioresour. Technol. 2003, 90, 285–295. DOI: 10.1016/s0960-8524(03)00133-0.
- Chua, K. J.; Mujumdar, A. S.; Hawlader, M. N. A.; Chou, S. K.; Ho, J. C. Convective Drying of Agricultural Products. Effect of Continuous and Stepwise Change in Drying Air Temperature. Drying Technol. 2001, 19, 1949–1960.
- Martynenko, A. Computer-Vision System for Control of Drying Processes. Drying Technol. 2006, 24, 879–888. DOI: 10.1080/07373930600734067.
- Raut, S.; Saleh, R. M.; Kirchhofer, P.; Kulig, B.; Hensel, O.; Sturm, B. Investigating the Effect of Different Drying Strategies on the Quality Parameters of Daucus Carota L. using Dynamic Process Control and Measurement Techniques. Food Bioprocess Technol. 2021, 14, 1067–1088. DOI: 10.1007/s11947-021-02609-y.
- Sturm, B.; Nunez Vega, A. M.; Hofacker, W. C. Influence of Process Control Strategies on Drying Kinetics, Colour and Shrinkage of Air Dried Apples. Appl. Therm. Eng. 2014, 62, 455–460. DOI: 10.1016/j.applthermaleng.2013.09.056.
- Winiczenko, R.; Górnicki, K.; Kaleta, A.; Martynenko, A.; Janaszek-Mańkowska, M.; Trajer, J. Multi-Objective Optimization of Convective Drying of Apple Cubes. Comput. Electron. Agric. 2018, 145, 341–348. DOI: 10.1016/j.compag.2018.01.006.
- Zhang, W.; Wang, K.; Chen, C. Artificial Neural Network Assisted Multiobjective Optimization of Postharvest Blanching and Drying of Blueberries. Foods 2022, 11, 3347. DOI: 10.3390/foods11213347.
- Hosseinpour, S.; Martynenko, A. An Adaptive Fuzzy Logic Controller for Intelligent Drying. Drying Technol. 2023, 41, 1110–1132. DOI: 10.1080/07373937.2022.2119996.