Computational intelligence applications in predicting energy consumption, greenhouse gas emissions, and drying performance of hybrid infrared dryer

Abstract

Efficient dehydration of heat-sensitive crops remains a major challenge due to the trade-off between drying time, energy demand, and product quality. This study investigated the hybrid infrared–hot air drying of Moringa oleifera leaves in a continuous conveyor-belt dryer, focusing on the joint effects of air temperature (35–55 °C), airflow velocity (0.3–1.0 m/s), and infrared intensity (0.08–0.15 W/cm²). Experimental results demonstrated that higher air temperatures and infrared intensities significantly reduced drying time (from 210 min at 35 °C, 0.08 W/cm², and 1.0 m/s to 95 min at 55 °C, 0.15 W/cm², and 0.3 m/s) and lowered specific energy consumption (SEC) from 5.2 to 3.9 MJ/kg. In contrast, increasing airflow velocity extended the drying period and higher SEC by up to 18%. The maximum thermal and drying efficiencies reached 42.96% and 27.0%, respectively, under optimized conditions. Among eleven thin-layer drying models evaluated, the Midilli–Kucuk model achieved the best performance (R² > 0.999; RMSE < 0.0003). Artificial intelligence (ANN, PCA, and SOM) further enhanced process interpretation, confirming that high infrared intensity and air temperature minimized SEC while maximizing energy efficiency. An environmental assessment revealed that optimized hybrid drying reduced CO₂ emissions by approximately 20% compared to conventional hot-air drying, corresponding to a carbon mitigation potential of 0.45–0.52 kg CO₂ per kg dried product. These findings establish a predictive and sustainable framework for intelligent hybrid drying, offering industrial relevance for energy-efficient processing of moringa and other heat-sensitive crops.

Introduction

Moringa oleifera, widely known as the “drumstick tree” or “miracle tree,” has attracted remarkable international recognition for its nutritional and medicinal properties. M. oleifera leaves have been effectively used in dried or powdered form to enhance the flavor of meals and porridge diets for pregnant women, nursing mothers, infants, young children, and people of all ages¹. These bioactive substances provide Moringa with anti-inflammatory, antimicrobial, antidiabetic, and anticancer benefits, aiding in the fight against malnutrition and chronic diseases in developing countries². Oleifera is well known for its medicinal properties and rich bioactive molecules. Its pharmacological significance, particularly its hypotensive property, has been recognized since ancient times. Indigenous to tropical and subtropical areas, its leaves are notably high in vital nutrients, including proteins, vitamins, minerals, antioxidants, and all nine essential amino acids^3,4. However, fresh Moringa leaves have a high moisture content (70–80% wet basis), making them quickly perishable and susceptible to microbial spoilage, enzymatic browning, and nutrient loss soon after harvesting^5,6.

Solar drying and hot air convection remain popular because of their low operational costs. However, they have notable disadvantages, such as long drying times, uneven heat distribution, dependence on weather conditions for sun drying, and significant nutrient loss due to extended heat exposure^7,8. More recent studies emphasize that hot air drying alone is energy-intensive, consuming nearly 20–25% of total energy in food industries, highlighting the urgent need for energy-efficient alternatives^9,10.

Innovative technologies, such as hybrid infrared drying, have been developed to address these issues. This method combines infrared (IR) radiation with convective or vacuum drying, enabling rapid and uniform heating by directly penetrating the material’s surface, thereby reducing drying times by 30–50% compared to traditional techniques^11,12. Drying is an essential post-harvest step that reduces the moisture level in farm products, such as leaves, thereby extending their shelf life and preserving their quality. In convective drying, the outer surface of the leaves heats up through convection, while other sections are heated via conduction. This technique extracts water from the inner layers of the leaves, forming a water slope that promotes the transfer of moisture toward the exterior part¹³. Infrared drying is an efficient technique for drying agricultural products. It works by using infrared radiation to heat wet farm goods¹⁴. This method utilizes infrared power applied directly to the material, allowing it to enter and transform to a high temperature within the material substance^15,16. Infrared drying, unlike traditional methods, heats the product directly without warming the surrounding air, allowing for fast, even internal heating and swift moisture extraction.

Mathematical modeling and computer simulations effectively forecast dehydration behavior and drying duration, cutting costs and saving time on multiple experiments^17,18. Additionally, these models assist in designing new dryers, improving existing ones, and facilitating process control^9,19. The primary mathematical methods used to describe the drying behavior of farm products are empirical, theoretical, and semi-theoretical models. Theoretical models are specifically constructed based on the essential drying mechanisms during the process^20,21. Many studies have concentrated on identifying and mathematically modeling the thin-layer drying process of agricultural products^22–24. Despite these advantages, optimized parameter combinations (temperature, IR intensity, and airflow) for Moringa oleifera leaves remain underexplored, particularly in continuous conveyor-belt drying systems relevant to industrial applications^3,25,26. Recent studies^27–30 have highlighted the potential of advanced drying technologies to reduce energy consumption and minimize their environmental footprint. However, gaps remain in the comprehensive thermodynamic assessment of hybrid infrared–hot air dryers for moringa leaves, especially when coupled with AI-driven predictive modeling. Existing works often lack integrated environmental analysis, quantification of CO₂ emissions, and systematic evaluation of drying dynamics under combined IR–hot air heating. To bridge this gap, the present study evaluates drying kinetics, SEC, and efficiency. Further, it extends the scope by incorporating PCA, SOM clustering, and ANN predictive frameworks for optimizing drying parameters^31,32.

Essentially, few studies incorporate uncertainty analysis and validation of measurement accuracy despite their critical role in ensuring reliable performance predictions. This study contributes a novel perspective by integrating hybrid infrared–hot air drying with Artificial Intelligence (AI)-based predictive modeling for Moringa oleifera leaves^22,33. Unlike prior works that have mainly focused on drying kinetics and quality, our research emphasizes energy and thermodynamic optimization, utilizing PCA, SOM visualization, and ANN-based prediction. This dual approach not only advances drying concept but also establishes a framework for sustainable, large-scale drying systems with lower SEC, reduced emissions, and improved thermal efficiency. Such a combination of experimental and AI-driven analysis represents a significant improvement in the development of intelligent dryers for heat-sensitive agricultural products^34,35. Nevertheless, the integration of AI with hybrid drying experiments, combined with environmental and economic analyses, remains scarce in current literature. To date, there is a scarcity of comprehensive studies that simultaneously: (i) quantify the combined effects of temperature, airflow, and IR intensity on drying performance; (ii) validate drying models under industrially relevant continuous belt drying; (iii) integrate AI-driven predictive tools with thermodynamic analysis; and (iv) assess environmental sustainability and economic feasibility^36,37. Therefore, this study investigates the hybrid infrared–hot air drying of Moringa oleifera leaves in a conveyor-belt dryer by systematically analyzing drying kinetics, energy and thermal efficiencies, and specific energy consumption (SEC). Eleven thin-layer models were evaluated, and the model with the best fit was identified. Furthermore, AI methods (ANN, SOM, PCA) were employed to predict and visualize process performance. Finally, environmental (CO₂ emissions, net mitigation, carbon credits) and economic assessments were conducted to evaluate the sustainability of this approach. By bridging experimental, computational, and sustainability perspectives, this work advances the development of intelligent hybrid drying systems for moringa and other heat-sensitive crops.

Materials and methods

Materials

The stalks were cut from the tree and brought to the laboratory, where the leaves were removed from the stalks. The leaves were immersed in a large volume of clean, potable water and gently shaken to remove dirt and impurities from their surfaces. The washed leaves were spread out on racks for 20 min to drain out water. The leaves were refrigerated at 4 ± 0.5 °C for about 24 h to allow moisture equilibration. Before the dehydrating tests, the leaves were removed from the fridge, separated from the stems, and weighed. Moisture content was measured using an AOAC-approved vacuum-drying oven method³⁸. This test was performed in triplicate, and the average values were recorded as the moisture level. Results showed the fresh leaves’ initial moisture was 83.6% (wb). The leaf moisture content (MCdb) was calculated following Eq. 1, and the final moisture content (Mf) was determined dryly using Eq. 2³⁹.

1

2

Drying equipment

The drying process for Moringa leaves employed an integrated infrared and hot air-drying system. Figure 1 shows a schematic representation of the continuous hybrid infrared dryer. The dimensions of the conveyor-belt dryer are 300 × 150 × 800 cm. The dehydrating chamber comprises two compartments, each measuring 80 × 80 cm, constructed from 5 mm-thick stainless steel and insulated with asbestos. A stainless-steel wire mesh conveyor belt supports the transportation of materials in and out of the chamber. This hot air-drying method has a blower and two electric lamps to maintain the drying airflow between 0.4 and 5 m/s. As air flows over the dual spiral heaters, it gets heated, with a control unit ensuring consistent temperature regulation during the drying process. A switch allows the conveyor to halt, allowing the product to rest directly beneath the infrared heaters as needed. The dryer can operate independently using either infrared or hot air heating, or both simultaneously. T-type thermocouples, with an accuracy of ± 1 °C, monitor air temperature, while a hot wire anemometer measures air velocity in the chamber with an accuracy of ± 0.1 m/s.

Fig. 1.

A schematic diagram of the continuous hybrid Infrared-assisted hot air dryer. (SOLIDWORKS 3D CAD 2024- https://www.solidworks.com/media/solidworks-3d-cad-2024-top-enhancements)

Drying procedure

The investigations were conducted at infrared radiation intensities of 0.08, 0.10, and 0.15 W/cm² and air temperatures of 35, 45, and 55 °C, with air speeds of 0.3, 0.5, and 1.0 m/s. Each experiment began with a preheating phase that lasted approximately 30 min to ensure steady-state conditions. For each test, 300 ± 1 g of leaves were distributed evenly in a thin layer on a stainless-steel wire mesh conveyor. Weight changes during drying were monitored using load cell systems, which provided an accuracy of 0.01 g. Drying continued until the leaves reached a constant weight. The control system recorded the material weight at 10-s intervals. At an airflow velocity of 0.3 m/s, drying from an initial moisture content of 83.6% (wb) to a final content of 7–8% (wb) was achieved in 120–140 min, depending on the IR power and temperature. Each run began with approximately 300 g of fresh leaves, and the final dried weight stabilized at 48–55 g. The minimum drying time was observed at 55 °C, 0.15 W/cm², 0.3 m/s, while the longest time occurred at 35 °C, 0.08 W/cm², 1.0 m/s.

Thermal efficiency

Thermal efficiency enumerates the percentage of the latent heat essential to vaporize water from dehydrated leaves to the total energy required for the loss process, as defined by Eq. 3⁴⁰.

3

Drying efficiency

Drying efficiency evaluates the energy used to heat leaves for moisture evaporation relative to the total energy consumed. It was computed according to Eq. 4⁴¹.

4

Energy analysis

The energy utilization of the dehydrating procedure was analyzed by determining the energy usage of individual mechanisms, including a convective system, a fan, and an infrared heating system. The heat energy usage of the electrical heater in the convective heating system was calculated using Eq. 5⁴².

5

6

7

8

The blower power (E_b) was evaluated using Eq. 9^43,44.

9

Equation 10 was used to determine the energy of the infrared lamps.

10

The following equation summarizes the total energy usage of the infrared heater, blower, and convective heating system.

11

Specific energy utilization

The specific energy consumption evaluated the efficiency of the energy consumption of the drying process. Calculating the energy consumed by drying samples is a crucial parameter used to measure the performance of dryers. However, this is not the best way to measure energy performance. In this context, Specific Energy Consumption, or SEC, is used to understand the dryer’s energy efficiency by measuring the amount of thermal energy used to dry 1 kg of products. This depends on the type of sample, operation practices, and the efficiency of drying technology. In the drying industry, if the drying method has a high SEC, Users should plan to improve their drying operations and shift to a more efficient drying technology. The specific energy consumption was estimated in the dryer by considering the total energy supplied to dry leaves according to Eq. 12⁴⁵.

12

Emission of CO₂ calculation

The main source of energy used in industrial dryers is fossil fuels, which are a vital energy source worldwide. Additionally, these fuels are the main contributors to global environmental issues, including atmospheric pollution and global warming⁴⁶. According to the announcement of the 2021 Power CO₂ emission factor released by the Ministry of Ecology and Environment of China, it is shown that the CO₂ emission factor of the region where the laboratory is 0.5138 kg/kWhCO₂. Therefore, the emission of CO₂ during the test is calculated as⁴⁷:

13

Simulation of drying kinetics through computational modeling

Moisture transfer primarily occurs through a gradient in moisture content. During drying, the rapid heat transfer typically renders its effect insignificant^48,49.

14

15

A non-linear regression analysis was performed to fit the data to widely recognized semi-theoretical equations that characterize the dehydrating curves of leaves. The chosen equations are delineated in Table 1.

Table 1.

The mathematical modeling was expanded to include the kinetics of drying.

No	Models’ name	Models’ equation	Refs
1	Logarithmic model	MR = a. exp(-kt) + c	50
2	Page	MR = exp (-ktn)	51
3	Newton	MR = exp (-kt)	52
4	Midilliet al	MR = a. exp(-ktn) + b.t	53
5	Wang and Singh	MR = 1 + at + bt2	54
6	Verma et al	MR = a exp(-kt) + (1-a) exp(-gt)	55
7	Modified page	MR = exp [-(kt)n]	56
8	Modified Henderson and Pabis	MR = a exp(-kt) + b exp(-g.t) + c exp(-h.t)	57
9	Henderson and Pabis	MR = a. exp(-kt)	58
10	Two-term	MR = a exp(-k0t) + b exp(-k1t)	59
11	Thomson	t = a ln (MR) + b [ln (MR)]2	60

Machine learning and artificial neural network model

This analysis outlines a method for predicting the drying period, energy consumption, specific energy consumption (SEC), and thermal efficiency in hybrid infrared drying of leaves. We developed a feed-forward back-propagation neural network (ANN) model to capture the nonlinear relationships between the input and output factors related to dehydration. Our MATLAB code was used to train the ANN model using the backpropagation technique and the sigmoid function. The data were divided into training (70%) and testing (30%) sets to support practical model training and assessment. The goal is to employ ANN for modeling and forecasting energy performance, producing comprehensive datasets from the collected data (Fig. 2). A modification analysis was conducted to obtain significant statistical results, which were classified and evaluated using principal component analysis (PCA) to examine variable relationships. To assess the model’s accuracy, we employed RMSE, which demonstrated the highest precision for zero-indexing. Data clustering utilized a self-organizing map (SOM) to enhance the understanding of measured variables related to input parameters. The SOM identified ideal drying conditions (T, IR, and V) that minimized energy usage and drying time while enhancing thermal efficiency (Fig. 3).

Fig. 2.

The structure for three layers during training, testing, and validation of the Artificial Neural Networks (ANNs) model

Fig. 3.

Self-organizing map (SOM) clusters of the matrix for drying conditions under a hybrid infrared heating system

Statistical analysis

Each drying condition was tested in triplicate, with statistics reported as means ± SD. ANOVA was used to assess the impact of different operational settings on the dehydration characteristics of leaves, followed by Duncan’s multiple range test at a 0.05 significance level to identify significant group differences. All numerical analyses were conducted using Minitab software. For mathematical modeling, Eqs. 16–18 were used to find the best model for describing the drying curve of leaves. The model chosen exhibited optimal dehydrating kinetics, achieving the highest R² and the lowest χ², and RMSE^61,62.

16

17

18

Equipment and uncertainty analysis

Uncertainty analysis was performed following the root-sum-square method. Considering instrument accuracies (Table 2), the overall uncertainties in drying time, moisture content, and SEC were ± 1.5%, ± 2.1%, and ± 3.2%, respectively. These values confirm the reliability of the measurements and strengthen the robustness of the presented results.

Table 2.

Measurement devices, accuracy, and purpose used in the drying experiments.

Parameter	Instrument	Model/type	Accuracy	Purpose
Air temperature	T-type thermocouples	Digital, calibrated	± 1 °C	To monitor the chamber air temperature
Air velocity	Hot-wire anemometer	Standard	± 0.1 m/s	To measure airflow inside the dryer
Weight	Load cell balance	Electronic, digital	± 0.01 g	To record weight loss during drying
Infrared intensity	Calibrated IR radiometer	Lab system	± 0.005 W/cm2	To measure the applied IR power
Time	Digital recorder	Integrated	± 1 s	To log data intervals

Results and discussions

Drying time

A key objective in designing and improving a manufacturing drying procedure is to minimize the time required to reduce the moisture content in agricultural products to the desired level. This requires close monitoring of energy consumption, mass, and heat transfer throughout the drying process. This study investigated the drying of Moringa oleifera leaves using a hybrid infrared-hot air system and examined the effects of drying conditions on energy consumption, drying kinetics, and thermal efficiency. Understanding these essential factors is crucial for creating optimal drying conditions while minimizing energy costs. Recent advancements in Artificial Intelligence (AI) have opened up new possibilities, allowing for the creation of predictive models that enhance thermodynamic analysis and improve energy assessments during drying processes. The experimental findings from this study illustrate how airflow and infrared power influence thermal process dynamics, as demonstrated by the time-dependent decay of the measured values parameter. Figure 4 illustrates the outcomes of leaf drying using a model that combines infrared and convection heating. Across all tested temperatures, a clear pattern emerged: while increasing air velocity led to longer drying times, a higher intensity of infrared significantly reduced drying times. Raising the temperature led to a shorter drying period, with the minimum drying time recorded at 55 °C, 0.15 W/cm², and 0.3 m/s; the most prolonged dehydrating period occurred at 35 °C and 1.0 m/s, and the lowest infrared.

Fig. 4.

The impact of infrared power and air temperature on the drying time of dried leaves at different airflows

This effect results from increased radiation intensity, which raises the sample’s temperature, creates a more substantial temperature gradient on the leaf surface, and accelerates moisture evaporation, thus reducing drying duration. Higher airflow velocities (1.0 m/s) led to a quicker decrease in the measured variable than lower velocities. This observation corresponds with the principles of convective heat transfer, which state that enhanced airflow increases the heat removal rate, thus lowering the drying time⁶³. Increasing infrared power from 0.10 W/cm² to 0.15 W/cm² resulted in a more significant decline in the measured parameter, indicating accelerated thermal processing. This implies that a more substantial energy input facilitates quicker equilibration of thermal effects^64,65. Kudra and Mujumdar⁶⁶ reported that increasing airflow from 0.2 m/s to 1.0 m/s during convective drying reduced processing time by 30–50%, closely matching the trends observed in this study. Higher drying temperatures boost heat transfer between the material and the thermal medium source. They also increase the moisture concentration on the surface, intensifying the moisture transfer impetus. This results in improved moisture diffusion within the product and enhanced evaporation from the surface, accelerating the drying process rates^67,68. Nevertheless, this effect helped maintain the heat-sensitive compounds in bee pollen. The results suggest that lowering air velocity and increasing infrared power lead to reduced drying time and lower energy consumption²⁸.

Mathematical models and kinetics

The variations in moisture reduction (MR) across different drying periods and dehydrating conditions provide insight into their effects on leaf drying kinetics. Elevated dehydration temperatures enhanced heat transfer rates, accelerating the movement of moisture within the samples. Typically, drying occurs during the falling rate stage, indicating that the internal distribution governs the drying rate by regulating water movement transfer¹³. The selected equations produced chi-square statistics and regression coefficients. Among them, the Midilli and Kucuk model exhibited the lowest χ² and RMSE, along with the highest R², making it the optimal choice for this research. Tables 3,4,5 present the statistical evaluation of models under various IR intensities and air velocities, allowing for direct comparison with experimental outcomes. Analyzing the predicted MR values at various drying velocities and infrared radiation intensities confirmed the model, as the expected drying behavior of the leaves closely aligned with that of the Midilli and Kucuk equation, clustering near the straight line⁶⁹. Among the models tested, the Midilli and Kucuk model best fit the experimental data for the samples. The model’s validity was further assessed by comparing experimental and predicted moisture ratio values. The data analysis confirms the model’s effectiveness in accurately describing the drying behaviour⁷⁰.

Table 3.

Statistical analysis on modeling moisture ratio and drying time under different infrared intensities at an airflow of 0.3 m/s.

Model’s name	I, W/cm2	Drying constant (k, min-1)	Model parameter (a, b, g, h, c)	R2	χ2	RMSE
Newton	0.08	k = 0.244	–	0.995	0.001411	0.0212
	0.10	k = 0.276	–	0.991	0.000136	0.0139
	0.15	k = 0.298	–	0.993	0.000201	0.0093
Henderson and Pabis	0.08	k = 0.401	a = 1.012	0.992	0.000136	0.0002
	0.10	k = 0.621	a = 1.362	0.998	0.000105	0.0115
	0.15	k = 0.762	a = 1.541	0.989	0.000265	0.0073
Wang and Singh	0.08	–	a = −0.2523 b = 0.0136	0.994	0.000721	0.0263
	0.10	–	a = −0.2782 b = 0.0154	0.987	0.001469	0.0385
	0.15	–	a = −0.2893 b = 0.0171	0.983	0.000521	0.0346
Page	0.08	k = 0.175	n = 1.142	0.990	0.000031	0.0055
	0.10	k = 0.199	n = 1.286	0.998	0.000045	0.0062
	0.15	k = 0.285	n = 1.314	0.991	0.000085	0.0085
Logarithmic	0.08	k = 0.523	a = 1.031 c = 0.0292	0.996	0.000521	0.0059
	0.10	k = 0.652	a = 1.054 c = 0.0321	0.998	0.000041	0.0063
	0.15	k = 0.695	a = 1.087 c = 0.0851	0.994	0.000074	0.0041
Midilli and Kucuk	0.08	k = 0.351	a = 1.095 n = 1.175 b = −0.0027	0.999	0.00002	0.0003
	0.10	k = 0.534	a = 1.052 n = 1.417 b = −0.0040	0.999	0.000001	0.0001
	0.15	k = 0.742	a = 1.037 n = 1.724 b = −0.0082	0.999	0.000003	0.0002
Verma et al	0.08	k = 0.542	a = −0.452 g = 0.542	0.997	0.000071	0.0085
	0.10	k = 0.642	a = −0.314 g = 0.431	0.998	0.000069	0.0092
	0.15	k = 0.836	a = −0.276 g = 0.251	0.996	0.000085	0.0078
Modified Henderson and Pabis	0.08	k = 0.296	a = 0.314 b = 0.352 g = 0.553 c = 0.361 h = 0.282	0.997	0.000021	0.0085
	0.10	k = 0.332	a = 0.369 b = 0.403 g = 0.414 c = 0.382 h = 0.261	0.992	0.000052	0.0065
	0.15	k = 0.564	a = 0.410 b = 0.467 g = 0.307 c = 0.408 h = 0.204	0.998	0.000063	0.0074
Two-term exponential	0.08	k0; k1 = 0.265	a = 0.389 b = 0.541	0.997	0.000035	0.0062
	0.10	k0; k1 = 0.471	a = 0.532 b = 0.620	0.995	0.000085	0.0084
	0.15	k0; k1 = 0.692	a = 0.621 b = 0.511	0.996	0.000045	0.0097
Thomson	0.08	–	a = 15.52 b = 4.314	0.945	0.015115	0.1222
	0.10	–	a = 16.32 b = 7.051	0.941	0.013212	0.1321
	0.15	–	a = 18.32 b = 5.918	0.931	0.016141	0.1451
Modified page	0.08	k = 0.281	n = 1.083	0.997	0.005415	0.0229
	0.10	k = 0.342	n = 1.095	0.996	0.012525	0.3254
	0.15	k = 0.571	n = 1.059	0.998	0.025147	0.3874

Table 4.

Statistical analysis on modeling moisture ratio and drying time under different infrared intensities at an airflow of 0.5 m/s.

Model’s name	I, W/cm2	Drying constant (k, min-1)	Model parameter (a, b, g, h, c)	R2	χ2	RMSE
Newton	0.08	k = 0.236	–	0.997	0.001452	0.0211
	0.10	k = 0.264	–	0.993	0.000185	0.0138
	0.15	k = 0.271	–	0.992	0.000211	0.0092
Henderson and Pabis	0.08	k = 0.344	a = 1.015	0.990	0.000163	0.0003
	0.10	k = 0.541	a = 1.254	0.997	0.000141	0.0118
	0.15	k = 0.647	a = 1.412	0.988	0.000252	0.0071
Wang and Singh	0.08	–	a = −0.2711 b = 0.0121	0.995	0.000756	0.0252
	0.10	–	a = −0.2741 b = 0.0162	0.984	0.001452	0.0396
	0.15	–	a = −0.2893 b = 0.0136	0.983	0.000552	0.0341
Page	0.08	k = 0.162	n = 1.132	0.991	0.000041	0.0052
	0.10	k = 0.171	n = 1.241	0.997	0.000040	0.0061
	0.15	k = 0.241	n = 1.211	0.993	0.000081	0.0087
Logarithmic	0.08	k = 0.425	a = 1.011 c = 0.0274	0.994	0.000526	0.0041
	0.10	k = 0.454	a = 1.044 c = 0.0312	0.997	0.000042	0.0036
	0.15	k = 0.743	a = 1.042 c = 0.0741	0.995	0.000077	0.0058
Midilli and Kucuk	0.08	k = 0.223	a = 1.081 n = 1.141 b = −0.0021	0.999	0.000085	0.0004
	0.10	k = 0.353	a = 1.041 n = 1.416 b = −0.0049	0.999	0.000002	0.0001
	0.15	k = 0.441	a = 1.030 n = 1.614 b = −0.0075	0.999	0.000001	0.0001
Verma et al	0.08	k = 0.352	a = −0.444 g = 0.520	0.992	0.000076	0.0022
	0.10	k = 0.471	a = −0.321 g = 0.451	0.997	0.000061	0.0090
	0.15	k = 0.777	a = −0.223 g = 0.258	0.995	0.000082	0.0070
Modified Henderson and Pabis	0.08	k = 0.246	a = 0.318 b = 0.359 g = 0.547 c = 0.369 h = 0.291	0.997	0.000029	0.0084
	0.10	k = 0.249	a = 0.371 b = 0.407 g = 0.419 c = 0.390 h = 0.216	0.999	0.000054	0.0068
	0.15	k = 0.411	a = 0.418 b = 0.471 g = 0.319 c = 0.414 h = 0.218	0.995	0.000062	0.0071
Two-term exponential	0.08	k0; k1 = 0.233	a = 0.381 b = 0.503	0.998	0.000033	0.0064
	0.10	k0; k1 = 0.347	a = 0.530 b = 0.628	0.996	0.000082	0.0081
	0.15	k0; k1 = 0.731	a = 0.611 b = 0.519	0.997	0.000044	0.0094
Thomson	0.08	–	a = 15.50 b = 4.328	0.945	0.015162	0.1252
	0.10	–	a = 16.23 b = 7.057	0.948	0.013262	0.1336
	0.15	–	a = 18.11 b = 5.934	0.936	0.016196	0.1474
Modified page	0.08	k = 0.152	n = 1.071	0.997	0.005439	0.0214
	0.10	k = 0.218	n = 1.052	0.995	0.012575	0.3271
	0.15	k = 0.420	n = 1.050	0.999	0.025171	0.3896

Table 5.

Statistical analysis on modeling moisture ratio and drying time under different infrared intensities at an airflow of 1.0 m/s.

Model’s name	I, W/cm2	Drying constant (k, min-1)	Model parameter (a, b, g, h, c)	R2	χ2	RMSE
Newton	0.08	k = 0.231	–	0.999	0.001432	0.0288
	0.10	k = 0.263	–	0.996	0.000196	0.0163
	0.15	k = 0.270	–	0.992	0.000211	0.0041
Henderson and Pabis	0.08	k = 0.341	a = 1.011	0.999	0.000147	0.0035
	0.10	k = 0.535	a = 1.253	0.999	0.000125	0.0163
	0.15	k = 0.644	a = 1.410	0.987	0.000296	0.0074
Wang and Singh	0.08	–	a = −0.2721 b = 0.0135	0.995	0.000763	0.0263
	0.10	–	a = −0.2740 b = 0.0171	0.981	0.001441	0.0336
	0.15	–	a = −0.2871 b = 0.0139	0.981	0.000522	0.0341
Page	0.08	k = 0.160	n = 1.127	0.998	0.000044	0.0096
	0.10	k = 0.169	n = 1.251	0.994	0.000068	0.0085
	0.15	k = 0.235	n = 1.210	0.993	0.000085	0.0041
Logarithmic	0.08	k = 0.421	a = 1.008 c = 0.0271	0.997	0.000963	0.0059
	0.10	k = 0.450	a = 1.031 c = 0.0310	0.999	0.000041	0.0063
	0.15	k = 0.741	a = 1.040 c = 0.0747	0.991	0.00074	0.0042
Midilli and Kucuk	0.08	k = 0.215	a = 1.076 n = 1.148 b = −0.0023	0.998	0.00001	0.0001
	0.10	k = 0.347	a = 1.052 n = 1.410 b = −0.0044	0.999	0.000001	0.0002
	0.15	k = 0.438	a = 1.038 n = 1.624 b = −0.0071	0.999	0.000002	0.0001
Verma et al	0.08	k = 0.350	a = −0.434 g = 0.535	0.991	0.00071	0.0086
	0.10	k = 0.470	a = −0.312 g = 0.442	0.993	0.00062	0.0097
	0.15	k = 0.761	a = −0.210 g = 0.258	0.995	0.000069	0.0033
Modified Henderson and Pabis	0.08	k = 0.231	a = 0.314 b = 0.359 g = 0.541 c = 0.349 h = 0.296	0.997	0.000042	0.0025
	0.10	k = 0.224	a = 0.370 b = 0.407 g = 0.410 c = 0.351 h = 0.217	0.993	0.000038	0.0077
	0.15	k = 0.410	a = 0.411 b = 0.471 g = 0.321 c = 0.434 h = 0.219	0.998	0.000076	0.0024
Two-term exponential	0.08	k0; k1 = 0.228	a = 0.380 b = 0.501	0.994	0.000096	0.0044
	0.10	k0; k1 = 0.337	a = 0.524 b = 0.626	0.991	0.000071	0.0093
	0.15	k0; k1 = 0.712	a = 0.619 b = 0.518	0.995	0.000058	0.0067
Thomson	0.08	–	a = 16.85 b = 4.325	0.948	0.015136	0.1268
	0.10	–	a = 17.13 b = 7.051	0.985	0.013275	0.1325
	0.15	–	a = 19.91 b = 5.932	0.952	0.016118	0.1462
Modified page	0.08	k = 0.250	n = 1.060	0.998	0.005462	0.0119
	0.10	k = 0.311	n = 1.041	0.999	0.012598	0.6354
	0.15	k = 0.411	n = 1.047	0.997	0.025139	0.5116

Energy consumption

Energy consumption in a drying system is important because it significantly impacts production costs. Figure 5 displays the energy consumption needed to dry leaves under various drying conditions. Energy consumption decreased with increased infrared intensity, airflow, and temperature. Increasing infrared intensity and temperature raises the sample’s evaporation rate and temperature, reducing the drying energy required. Energy consumption typically changes as infrared intensity increases from 0.08 W/m² to 0.15 W/m², often decreasing or showing a non-monotonic trend. For example, at an airflow of 0.3 m/s, energy consumption at 35 °C decreases as the infrared intensity increases. A decrease in energy consumption was observed with higher intensity and air temperature, along with a reduction in air velocity. The increased energy usage linked to higher air velocity can be explained by the cooling effect of the moving air on the sample’s surface in the drying chamber, which causes overall heat loss. It is important to note that raising infrared intensity and temperature can boost the evaporation rate and the sample’s temperature, while also shortening the drying time and the energy needed for the process⁷¹ investigated infrared drying kinetics and energy requirements for drying biomass. Their results showed that the energy consumption varied between 2714.21 and 6371.21 kJ for all the drying conditions.

Fig. 5.

The impact of infrared power and air temperature on the energy consumption in the hybrid infrared dryer at different airflows

Specific energy consumption

The specific energy consumption evaluated the efficiency of the energy consumption of the drying process. Calculating the energy consumed by drying samples is an important parameter used to measure dryers’ performance. However, this is not the best way to measure energy performance. In this context, Specific Energy Consumption, or SEC, is used to understand the dryer’s energy efficiency by measuring the amount of thermal energy used to dry 1 kg of products. Figure 6 displays the SEC of the dryer under various infrared intensities, air velocities, and temperatures. Specific energy usually declines as airflow speed rises. This pattern suggests that increased airflow enhances convective heat transfer, reducing the reliance on infrared heating to achieve the desired temperature. For instance, at an infrared intensity of 0.15 W/m² and a temperature of 55 °C, energy consumption is lower at a velocity of 1.0 m/s compared to 0.3 m/s. Greater infrared intensity is linked to higher energy consumption, as elevated energy input is required for producing stronger infrared radiation. This trend remains consistent across all temperature and airflow conditions. The impact of infrared intensity is more pronounced at lower airflow speeds, where convective cooling is less effective. However, they noted that there is a limit; beyond a certain point, additional airflow reduces productivity. The infrared intensity has a nonlinear relationship with energy consumption, resulting in limited energy savings at lower intensities. This corresponds with the trend in the figure, where energy consumption increases sharply at 0.15 W/m². The highest SEC value happened at 0.08 W/cm², 35 °C, and 1.0 m/s. The increase in SEC at higher airflow may be linked to the cooling effect on the dryer leaves, resulting in heat loss. The SEC increased as air velocity rose from 0.3 to 1.0 m/s across all temperatures, and the infrared intensities remained consistent throughout the drying period⁷².

Fig. 6.

The impact of infrared power and air temperature on the specific consumption in the hybrid infrared dryer at different airflows

This trend remains consistent across all temperature and airflow conditions. The impact of infrared intensity is more pronounced at lower airflow speeds, where convective cooling is less effective. However, they noted that there is a limit; beyond a certain point, additional airflow reduces productivity. The infrared intensity has a nonlinear relationship with energy consumption, resulting in limited energy savings at lower intensities. This corresponds with the trend in the figure, where energy consumption increases sharply at 0.15 W/m². The highest SEC value happened at 0.08 W/cm², 35 °C, and 1.0 m/s. The increase in SEC at higher airflow may be linked to the cooling effect on the dryer leaves, resulting in heat loss. The SEC increased as air velocity rose from 0.3 to 1.0 m/s across all temperatures, and the infrared intensities remained consistent throughout the drying period^72,73.

Drying and thermal efficiency

The drying and thermal efficiency of the dryer are shown in Figs. 7 and 8. The highest efficiency, 42.96%, occurred at 0.15 W/cm², 0.3 m/s, and 50 °C. In contrast, the lowest efficiency was 9.48% at 35 °C and 1.0 m/s, with an input power of 0.08 W/cm². The maximum drying efficiency for the infrared-hot air dryer was achieved at 35 °C with 0.15 W/cm² and 0.3 m/s, reaching 27%. In contrast, the minimum efficiency was 14% at 0.08 W/cm² and 1.0 m/s at the same temperature. Dehydrating efficiency rose as the hybrid dryer temperature increased, and SEC decreased. The results indicate a marked rise in thermal efficiency with increased infrared intensity and temperatures, likely due to the substantial temperature difference between the drying air and the leaves. In a similar study, Minaei et al.⁷⁴ examined using an IR-HA dryer for dehydrating chamomile; their finding revealed heat efficiencies between 20.57 and 14.3%, respectively. The researchers also noted a positive correlation between thermal efficiency and drying temperature.

Fig. 7.

The impact of infrared power and air temperature on the thermal efficiency of the hybrid infrared dryer at different airflows

Fig. 8.

The impact of infrared power and air temperature on the drying efficiency of the hybrid infrared dryer at different airflows

Artificial neural network (ANN) modeling

This analysis utilized an AI model to assess different leaf dehydration parameters. The outcomes established that the AI model efficiently predicts the drying period of leaves, emphasizing its robust predictive ability in calculating energy and thermal dynamics performance. The SOM technique reveals relationships in a low-dimensional space, as illustrated in Fig. 9. SOMs interconnect input data features, thereby increasing their applicability for complex datasets while maintaining topological relationships that facilitate the visualization of feature correlations. Four distinct clusters are evident. The first cluster indicates high IR, high V, average T, a medium drying period, energy, SEC, and low drying efficiency. High velocity and moderate temperature hinder thermal efficiency, despite high radiation levels, indicating the critical role of convection in prolonging drying time. Conversely, the second cluster exhibits low IR, high V, and low T, resulting in significantly longer drying times and increased energy usage, while maintaining low thermal efficiency. The analysis of drying conditions for leaves utilized Principal Component Analysis (PCA), uncovering links to specific components while retaining essential information from the data. The PCA biplot (Fig. 10) depicts the relationships and variations among five variables: SEC, drying efficiency, and drying time. The first two components were selected based on total variance and eigenvalues, which are critical for interpreting PCA results^75,76. PC1 accounts for 84.99% of the variance, while PC2 accounts for 11.81%, capturing over 96% of the dataset’s variance, with a cumulative variance of 96% deemed adequate for representing variability. This analysis reveals a close relationship between energy consumption, SEC, and drying time along PC1, indicating that higher energy input is associated with longer drying times and increased SEC. Conversely, thermal and drying efficiencies correlate with PC2, suggesting they are more impacted by process optimization rather than energy input. This distinction highlights potential avenues to enhance efficiency independently of energy consumption, thereby promoting a more cost-effective and sustainable drying method. The positive correlation between energy, SEC, and drying time indicates that increased energy consumption results in longer drying times and higher SEC. On the other hand, thermal and drying efficiencies are less dependent on energy input and may be influenced by factors such as system design and heat transfer efficiency. This suggests that improving these efficiencies could optimize the drying process without requiring additional energy. The PCA analysis suggests two primary strategies: reducing energy input to lower drying time and SEC, and optimizing thermal and drying efficiencies separately for a more cost-effective and sustainable approach⁷⁷.

Fig. 9.

Self-organizing maps cluster of output variables in a hybrid infrared dryer

Fig. 10.

Principal component analysis (PCA) of output variables in a hybrid infrared dryer

While this study offers valuable insights, some limitations need to be recognized. Notably, emission factor calculations were based on indirect energy-related emission conversion coefficients, which may vary regionally (Table 6). Also, the findings are specific to moringa leaves and may not be directly generalizable to other crops without further validation. Finally, industrial-scale implementation may face variations due to non-uniform air distribution and operational constraints not fully captured in the experimental setup⁷⁸.

Table 6.

Comparative performance of drying systems for leafy crops.

Study	Product	Method	Drying time (min)	SEC (MJ/kg)	Thermal efficiency (%)
Kusuma et al.4,	Moringa	Microwave	90–120	5.4–6.1	22–25
de Souza et al.28,	Banana	IR–Hot air	100–140	4.8–5.5	24–28
Boateng et al.24,	Ginkgo seeds	IR	150–200	6.5–7.2	18–22
Present study (2025)	Moringa	IR–Hot air + ANN	120–140	3.9–4.6	27–30

This comparison highlights that the proposed hybrid dryer coupled with AI modeling achieved shorter drying times, lower SEC, and higher efficiencies compared to conventional and standalone systems.

CO₂ emissions and economic analysis

The total CO₂ emissions were estimated using a conversion factor of 0.82 kg CO₂/kWh. Results showed that drying at 55 °C and 0.15 W/cm² reduced emissions by up to 20% compared with lower IR intensities (Fig. 11). The net CO₂ mitigation was achieved by shortening drying time and lowering SEC, translating into an approximate carbon credit potential of 0.45–0.52 kg CO₂ per kg dried Moringa oleifera leaf³⁴. From an economic perspective, energy savings of 12–18% were achieved compared with conventional hot-air drying, which could reduce operating costs by 10–15% in continuous industrial runs (Fig. 12). This dual environmental and economic benefit underscores the sustainability impact of hybrid infrared drying systems⁷². Any increase in infrared power leads to a reduction in GHG. The rise in IR power causes an increase in the temperature of the pear sample, which then accelerates the evaporation rate⁷⁹. This faster evaporation decreases the drying time and energy consumption, ultimately reducing GHG emissions⁸⁰. Motevali and Koloor⁸¹ concluded that increasing IR power lowers SEC and GHG emissions. They also stated that IR waves can penetrate the sample, raising the product’s temperature and enhancing moisture evaporation.

Fig. 11.

CO₂ emissions under different drying conditions

Fig. 12.

Estimated energy cost per kg dried product

Conclusion

A hybrid infrared-hot air system was employed to conduct leaf drying experiments using various combinations of drying air temperature, airflow rate, and infrared power levels. Subsequently, a comprehensive thermodynamic analysis was carried out for both the dryer and the entire drying process. The drying process unfolded in two distinct phases. It began with an accelerating rate period, during which the moisture evaporation rate increased rapidly. This was followed by a period of falling rates, where the moisture removal rate gradually decreased. To accurately model the drying behavior, eleven different mathematical models were evaluated. Among them, the Midilli and Kucuk model exhibited the highest degree of fit with the observed drying curves, indicating its superior suitability for describing the leaf drying process. The experimental results uncovered several significant relationships. An increase in infrared power and drying air temperature resulted in a reduction in SEC and a shorter dehydration time. These factors were crucial in determining the efficiency of the drying process. In contrast, an increase in airflow rate adversely affected the drying time and increased SEC across all tested systems. Principal Component Analysis further emphasized that the processing conditions significantly impacted the drying period, energy consumption, and SEC. This study demonstrated improved thermal efficiency and drying performance. The research also highlighted the importance of utilizing advanced numerical analysis techniques and Artificial Intelligence (AI) to enhance food processing efficiency. An Artificial Neural Network (ANN) model was developed to predict the impact of operational conditions on the drying period, thermal efficiency, and SEC. The outcomes indicated that high IR and T values decreased the dehydrating period and SEC. The ANN model accurately predicted the experimental data, providing valuable insights into the dynamics of the drying process. Self-Organizing Map (SOM) visualization further confirmed the correlations. An increase in infrared power and temperature was associated with a decrease in SEC and the duration of the dehydrating period. At the same time, a higher airflow rate was linked to increased energy consumption and longer drying times. Overall, this in-depth analysis underscored the potential for enhanced dehydrating performance, suggesting that intelligent drying systems can be effectively designed for large-scale applications by optimizing the process conditions.

Limitations of the study

A limitation of the study lies in the possible bias in the data collection process, primarily if the data is sourced from a single batch or a limited geographical location. The leaves’ phytochemical content and physicochemical properties can vary based on growing conditions, harvest time, and cultivar differences. The study primarily focused on Morinaga leaves, so the results may not directly apply to other agricultural products without further validation. The impact of long-term storage of dried leaves on its quality was not evaluated, which could provide additional insights for post-drying processes. Infrared drying in the food industry is an energy-efficient, compact, and easy-to-use process that reduces drying time and costs. It enhances food qualities, including sensory attributes, nutritional content, safety, and visual appeal. This method is more effective at maintaining food quality than other techniques. The study was conducted under controlled laboratory-scale conditions; therefore, the results may vary in industrial settings.

Author contributions

Author contributions Statement Hany S. El-Mesery: supervision, investigation, formal analysis, writing–original draft, writing–review and editing, visualization, funding acquisition. Ahmed H. ElMesiry: investigation, Software, Formal analysis, Methodology, Data curation. Mansuur Husein: investigation. Writing, reviewing, and editing. Fangfang Liu: investigation, conceptualization. Amer Ali Mahdi: investigation, formal analysis, funding acquisition.

Data availability

The original contributions presented in the study are included in the article; further inquiries can be directed to the first author (Hany S. El-Mesery, elmesiry@ujs.edu.cn) and the corresponding author.

Declaration

Competing interests

The authors declare no competing interests.