Design Optimization of Cleaning Fan Blades for Rice Combine Harvesters: An Experimental and CFD Simulation Study

Abstract

Recent advancements in rice breeding have significantly increased production in China. However, high-yielding varieties require strong airflow for effective cleaning. Longitudinal-flow rice combine harvesters equipped with a centrifugal fan with four blades are widely used in China; however, these fans exhibit fluctuating cleaning performance and airflow maldistribution. To address these limitations, this study developed an innovative multi-blade cleaning fan design by incorporating the blade clocking effect, a concept not previously applied in centrifugal fans. To support the design process, the required airflow rates and reduction in static pressure were first analyzed. Based on these findings and fundamental fan design theory, three fan models were designed with blade clocking angles of 0°, 5.5°, and 10.5°, respectively. Three fan models were evaluated through computational fluid dynamics (CFD) simulations using a design of experiments approach based on Box–Behnken design response surface methodology to identify the optimal fan. The fan features a 10.5° clocking angle, meeting the airflow requirements for effective cleaning. In the test bench measurements, the setup with guide plate angles No. 1 and No. 2 at 32° and a fan speed of 1200 rpm was identified as optimal. The newly designed multi-blade cleaning fan overcomes the limitations of conventional four-blade designs, significantly enhancing airflow uniformity.

1. Introduction

Rice is one of the most important grain crops in China, with a planting area of around 30 million hectares and an annual yield exceeding 200 million tons [1,2,3]. According to statistics from the Ministry of Agriculture and Rural Affairs of the People’s Republic of China [4], there are 2328 cultivated rice varieties, and recent advancements in rice breeding have significantly contributed to increased rice production. Notably, a high-yielding variety known as “Super Rice”, Y Liangyou 900, has achieved a remarkable yield of 16,000 kg/ha per harvest. These rice varieties are characterized by high yield, grain moisture content exceeding 30%, and stem moisture content over 70% [5,6,7,8]. However, changes in the physical characteristics of these varieties have a significant effect on the performance of the combine harvester’s grain-cleaning process. Modern rice production increasingly demands intelligent machinery and equipment [9], among which the combine harvester plays a central role [10]. These machines integrate multiple harvesting operations, including cutting, collecting, threshing, separating grains from material other than grain (MOG), cleaning, and transporting the cleaned grains into a grain bin for temporary storage [11]. By 2024, approximately 85% of rice was harvested using combine harvesters in China, and substantial progress has been made in the design of key working components such as the conveyor [12], threshing and separation systems [13,14,15,16,17], cleaning units [18], and the chassis [19] as well as in the development of intelligent technologies [20]. The performance of a combine harvester is strongly influenced by the efficiency of its cleaning unit, which comprises the fan, grain pan, and sieve. The threshed rice output falls onto the grain pan and is carried to the rear of the sieves through the combine action of the sieves and the airflow. The fan generates airflow within the cleaning system to facilitate the separation of grain from MOG. In rice combine harvesters manufactured by Chinese companies, a centrifugal fan with four blades is commonly used to generate the required airflow. However, its cleaning performance often exhibits large fluctuation [1,21]. This is due to airflow maldistribution, and rapid velocity decay at the outlet of the centrifugal fan [9,22]. A key challenge associated with the centrifugal fan with four blades is airflow maldistribution, and increased drag, especially under high-throughput harvesting conditions. Numerous researchers have explored various approaches to address this issue. Key research efforts include the following two aspects:

(1) Optimizing the working parameters of the cleaning fan, such as fan speed, air distributor angle (guide plate angle(GPA)), and airflow direction, is essential for achieving efficient cleaning performance [1]. Researchers studied the variation in airflow within a tangential-longitudinal flow combine harvester through experiments involving various combinations of working parameters [23]. Another study experimentally identified the optimal combination of fan parameters to maximize cleaning performance [24]. In another study, the impact of airflow outlet opening, fan speed, and deflector angle on the cleaning system of a grain combine harvester was studied, determining an optimal parameter combination [25]. Furthermore, researchers examined the influence of centrifugal fan working parameters on airflow distribution in the cleaning shoe, considering variables such as the upper sieve airflow supply distance, fan speed, and blade number [26]. The scholars mentioned have optimized working parameters, which offer short-term improvements in fan performance; however, they did not address the challenge of achieving long-lasting results.

(2) Optimization of the fan’s structural design primarily focuses on modifying the fan geometry to improve the performance of the cleaning system in combine harvesters. These efforts include enhancing fan outlet airflow distribution by increasing the width of the fan inlet [27]; optimizing airflow velocity distribution by analyzing the blade cut-off location, cut-off radius, expansion angle, and impeller width [28]; and improving cleaning performance by adding an inlet to a cross-flow fan [29]. Other studies have examined the influence of volute design and blade number on centrifugal fan performance [30]; demonstrated that the optimized blade design could stabilize airflow and improve cleaning effectiveness [31]; developed multi-duct fan systems to improve internal airflow distribution [1]; and optimized fan blade geometry to enhance airflow uniformity [21]. The abovementioned scholars have focused on optimizing the fan’s structural design, which has several advantages, including improved airflow distribution, enhanced cleaning efficiency, and long-term performance stability. By modifying key geometric parameters such as inlet width, blade cut-off location, impeller width, and volute shape, they have achieved more uniform and stable airflow, resulting in better separation of grain from MOG. However, airflow maldistribution, rapid velocity decay at the outlet, and increased airflow drag under high-throughput harvesting conditions still present significant challenges to achieving optimal fan performance.

Although significant progress has been made in improving airflow distribution both at the fan outlet and inside the cleaning shoe through parameter and structural optimization of cleaning fans, current methods still struggle to ensure stable and uniform airflow under varying field conditions. Additionally, previous studies have not sufficiently explored how blade clocking and blade spacing configurations influence airflow velocity and pressure distribution at the fan outlet and inside the cleaning shoe. In centrifugal fans, performance is significantly influenced by the arrangement of the blades relative to the volute tongue, a phenomenon known as the clocking effect. While the clocking effect has been extensively studied in compressors and pumps, its impact on the performance of centrifugal fans, particularly those used in combine harvesters, remains underexplored.

Therefore, this study is unique in applying the blade-clocking concept to centrifugal fan design for the first time to address the limitations of existing fan systems. To achieve this, estimation of the required airflow rate and the reduction in static pressure were first analyzed. Then, an innovative multi-blade fan was designed and developed by incorporating blade clocking angles. The significance of the study lies in enhancing airflow velocity and improving pressure distribution uniformity to meet the demands of modern high-yield rice varieties. These advancements will contribute to efficient rice harvesting, supporting farmers by ensuring high-quality harvesting.

2. Materials and Methods

2.1. Baseline Design and Optimization Process for the New Fan Blade Design

This refers to the original fan configuration selected as a reference point for performance evaluation and subsequent optimization. The conventional centrifugal fan with four blades in a combine harvester, as shown in Figure 1b, was used as the baseline design for this study. This fan is characterized by airflow maldistribution at the outlets and within the cleaning shoe and high grain sieve losses and impurity ratios [1,21]. A new cleaning fan blade was designed based on the required total static pressure (P, Pa) and the desired airflow rate (Qv, m³/s).

To address the limitations of the conventional fan, a new multi-blade fan structure was designed and developed to enhance airflow performance and uniformity. The detailed design methodology and optimization workflow are illustrated in the flowchart presented in Figure 2. This approach aligns with standard fan design principles widely used in agricultural machinery.

The optimization process focused on improving airflow uniformity and pressure distribution by adjusting blade spacing and clocking angles. Parameter optimization enhanced fan performance in terms of airflow distribution and total pressure generation. In this study, parameters such as blade clocking angle, blade spacing, and fan speed were varied and analyzed using CFD simulations. A design of experiments (DoE) approach was employed to efficiently explore the influence of these parameters on fan performance. Performance indicators, including total pressure rise, uniformity of airflow at the fan outlet, and airflow rate, were used as evaluation criteria. Based on the simulation results, the most effective parameter combinations were identified, leading to the selection of an optimized fan design that demonstrated high aerodynamic performance compared to the baseline design.

2.2. Determination of the Reduction in Static Pressure (P)

The reduction in static pressure (P) represents the drag the fan must overcome to ensure sufficient airflow through the fluidized grain layer and sieve during cleaning operations. This directly influences the fan’s ability to generate the required airflow for grain separation. Determining the reduction in static pressure involves collecting and analyzing the spatial distribution of threshed throughput materials. Thus, a custom-made throughput collection box (770 mm wide and 1500 mm long) was manufactured, as shown in Figure 3a. During harvesting, the combine harvesters’ cutting width was maintained at 1.8–2.0 m, the stubble height was 15–20 cm, and each test run was 20 m.

The distribution of threshed material was divided into nine measurement zones (Zone 1–Zone 9), as shown in Figure 4. Zones Z₁, Z₂, and Z₃ represent the front region; Z₄, Z₅, and Z₆ the middle region; and Z₇, Z₈, and Z₉ the rear region of the sieve. This zonal division provided a comprehensive analysis of threshed materials in both the transverse and longitudinal directions, as shown in Figure 5b. The components in each zone were weighed using an electronic balance after separation. The collected data were thoroughly analyzed using MATLAB R2023a, and a 3D surface graph was generated to visualize the results.

The reduction in static pressure was analyzed using a 12 × 5 grid of measurement points as shown in Figure 5b, with each point covering an area of 150 mm × 150 mm. Molenda et al. [32] indicated that grain at medium moisture content exhibits relatively lower porosity, resulting in increased airflow drag. Based on the fluidized grain layer height, porosity, and density of fluidized grain, the reduction in static pressure (P) and airflow drag coefficient ($c_f$) can be calculated using Equations (1) and (2), respectively [33].

$$P={\displaystyle \frac{M }{\left(1-{\epsilon }_f\right){\rho }_{g}A}}\left({\rho }_g-{\rho }_a\right)\left(1-{\epsilon }_f\right)g$$

(1)

$$c_f={\displaystyle \frac{2P }{{\rho }_a{V_f}^2}}$$

(2)

where $M$ is the total grain mass present in a specific zone, kg; A is the corresponding sieve area in that zone, m²; ${\epsilon }_f$ is porosity of the grain layer, with 40% used in this case; ${\rho }_a$ is the air density, assumed to be 1.225 kg/m³; $V_f$ is the velocity required to achieve fluidization, m/s; ${\rho }_g$ is the grain bulk density, taken as 1350 kg/m³; g is the gravitational acceleration, 9.81 m/s².

2.3. Estimation of the Required Airflow Rate (Q_v)

The required airflow rate (Q_v, in m³/s) was calculated to optimize the fan blade design for high-yielding rice varieties. A feed rate of 7 kg/s was used as the design reference. The total required airflow rate (Q_v) was computed using Equation (4) [33].

$$F=w_{c}vp\left(1+\gamma \right)$$

(3)

$$Q_v={\displaystyle \frac{\beta F}{\mu \rho }}$$

(4)

where Q_v is the required airflow rate, m³/s; F is feed rate (grain + MOG), kg/s; $\beta$ is the ratio of the MOG in the total threshed outputs (typically 10–15%); $\mu$ is mixed concentration ratio of straw carried in airflow is 0.2–0.3; $\rho$ is the airflow density, 1.225 kg/m³; w_c is the cutting width of combine harvester, 1.8 to 2.0 m; v is forward speed of combine harvester, 1.0–1.2 m/s; γ is the straw-to-grain ratio, 2.0–2.2; p is the grain mass per area, 1.02 kg/m².

2.4. New Fan Blade Design Procedure

2.4.1. Fan Type Selection

The right fan type was selected based on the calculated specific speed (N_s). In fan selection, the type of fan, namely centrifugal, axial, or mixed-flow, is typically determined based on the calculated specific speed ($N_s$) using Equation (5).

$$N_s={\displaystyle \frac{N\sqrt{Q_v}}{\sqrt[4]{P^3 }}}$$

(5)

where N_s is a non-dimensional specific speed; Q_v is the airflow rate, m³/s; P is the total static pressure reduction, Pa; N is the design fan speed, taken 1200 rpm.

2.4.2. Estimation of Pressure (ψ) and Flow (ϕ) Coefficients

The pressure coefficient (ψ) and flow coefficient (ϕ) are key non-dimensional parameters for evaluating fan performance. They help to determine whether the optimized fan meets airflow and pressure requirements for efficient cleaning, depending on the specified cleaning capacity. These coefficients are calculated using Equations (6) and (7) [34].

$$\psi ={\displaystyle \frac{2P}{{\rho }_a\times u_2^2}}$$

(6)

$$\varphi ={\displaystyle \frac{{4Q}_v}{\pi \times u_2{\times D}_2^2}}$$

(7)

$$u_2={\displaystyle \frac{\pi D_{2}N}{60}}$$

(8)

where ψ is the pressure coefficient; ϕ is the flow coefficient; u₂ is the tangential velocity, m/s; ${\rho }_a$ is the airflow density, 1.225 kg/m³; D₂ is the impeller outlet diameter, m.

2.4.3. Impeller Inlet Diameter and Blade Width Calculation

The impeller inlet diameter and blade width were calculated based on the required airflow rate (Q_v) and static pressure reduction (P) to meet the necessary airflow velocity for effective grain cleaning. The parameter b, as shown in Figure 6, represents the blade width and can be calculated using Equation (10) [35]. The fan inlet diameter (D₁), as shown in Figure 7b, is calculated using Equation (9).

$$D_1=k_0\sqrt[3]{{\displaystyle \frac{Q_v}{N-\left(1-\psi \right)\varphi }}}$$

(9)

$$b^{\prime }={\displaystyle \frac{D_1}{4}}\left(1-{\mathit{\Phi }}^2\right)\psi$$

(10)

where D₁ is the impeller inlet diameter, m; N is the fan speed, rpm; $k_0$ is the static pressure coefficient for blade angle < 90, 0.71; ψ is the pressure coefficient; ϕ is the flow coefficient; Φ is the ratio of the hub diameter (D₀) to the impeller outer diameter (D₂).

The required total width can be achieved by arranging three identical fans assembled side by side in a parallel configuration, ensuring uniform airflow distribution across the entire span of the cleaning system, as shown in Figure 8b.

2.4.4. Design of Fan Model Based on Clocking Effect

In this study, the blade arrangement was carefully modified by shifting the angular position of blades to 0°, 5.5°, and 10.5° relative to the volute tongue, as shown in Figure 9a–c. The fan models were designed using SolidWorks^® 2023 (Dassault Systèmes S.A., Waltham, MA, USA) software to ensure proper integration with the existing volute and to prepare for CFD simulation. The effects of blade clocking angles, blade spacing, and fan speed on the aerodynamic performance of fans were analyzed using CFD simulation.

2.5. Optimal Fan Model Selection Method Based on CFD Simulation

In this study, optimization parameters such as blade clocking angle, blade spacing, and fan speed were analyzed. The fan models were evaluated through CFD simulations under various DoE approaches based on the BBD-RSM to identify the optimal fan design. The CFD-based optimal fan selection involved three levels of blade clocking angle (δ = 0°, 5.5°, and 10.5°), three levels of blade spacing (t = 15 mm, 31 mm, and 48 mm), and three levels of fan speed (N = 1000, 1100, and 1200 rpm), resulting in a total of twelve conditions. The selected fan speeds (1000, 1100, and 1200 rpm) represent typical operating conditions in rice combine harvesters, allowing for the evaluation of fan performance across a practical working range [24]. The JMP Pro 17 software (SAS Institute Inc., Cary, NC, USA) was used to analyze the effects of the main parameters on maximum airflow velocity at fan outlets. p-values were quantified to determine the statistical significance.

2.5.1. Geometric Model, Mesh Generation, and Boundary Conditions

For CFD simulation, three different geometric models with clocking angles of 0°, 5.5°, and 10.5° were designed using SolidWorks^® 2023 (Dassault Systèmes S.A., Waltham, MA, USA) design software, as shown in Figure 10a,b. To improve computational efficiency, only one-half of the domain width was modeled, taking advantage of the fan’s symmetric structure [29]. The meshing process was conducted using ANSYS ICEMCFD 2021 R1 (ANSYS, Inc., Canonsburg, PA, USA), employing a tetrahedral mesh to capture the three-dimensional turbulent flow features. A transient simulation with a rotating reference frame was enabled through a sliding mesh strategy. The final mesh included approximately 3.76 million cells, as presented in Figure 10c,d.

2.5.2. Grid Independence

Grid independence was evaluated by gradually refining the mesh, with the total number of cells ranging from 2,280,770 to 3,757,268. The results showed that the flow coefficients obtained from the finest grids did not vary significantly, indicating that grid independence had been achieved. Therefore, to balance accuracy and computational efficiency, the mesh with 3,757,268 cells was selected for the simulations.

2.5.3. Governing Equations, Numerical Scheme, and CFD Simulation Setup

The intricate geometry of the fan blades made numerical convergence difficult. To overcome this, the simulation process began with steady-state calculations using MRF and FOD schemes. Once stable initial conditions were reached, the analysis transitioned to a transient regime. The CFD simulations were based on the Reynolds-averaged Navier–Stokes (RANS) equations, including the continuity and momentum equations. The RNG k–ε model is well-suited for capturing the transient flow characteristics in blade channels [36,37]. The continuum fluid field is calculated from the continuity and Navier–Stokes equations based on the local mean variables over a computational cell, which are given by

$${\displaystyle \frac{\partial \rho }{\partial t}}+{\displaystyle \frac{\partial \rho }{\partial x_i}}\left(\rho u_i\right)=0$$

(11)

$${\displaystyle \frac{\partial }{\partial t}}\left(\rho u_i\right)+{\displaystyle \frac{\partial \rho }{\partial x_j}}\left(\rho u_i{u}_j\right)={\displaystyle \frac{\partial p}{\partial x_i}}+{\displaystyle \frac{\partial }{\partial x_i}}(\mu {\displaystyle \frac{\partial u_i}{\partial x_j}})-\rho \overline{\acute{x_i}\acute{x_j}}+S_i$$

(12)

where $\rho$ is the density of air, 1.225 kg/m³;$u_i{ \mathrm{and} u}_j$ are the airflow velocity components, m/s; p is the pressure, Pa; $\mu$ is the dynamic viscosity, Pa/s; $g$ is the gravitational acceleration, 9.81 m/s²; $S$ is the momentum sink, kg m/s; $\acute{x_i}\acute{x_j}$ are the Reynolds stress terms, kg/m s².

To evaluate the aerodynamic performance, simulations were executed at a fan speed of 1200 rpm, with both guide plates No. 1 and guide plate No. 2 adjusted to an angle of 32°. The fan blade was modeled with a rotating wall boundary, while the stationary components were assigned no-slip conditions. The SIMPLE algorithm facilitated pressure–velocity, coupling with convergence set at a residual level of 1 × 10⁻⁴. Pressure boundaries were imposed at the inlet and outlet, both defined with zero gauge pressure. A transient approach was adopted, comprising 250 time intervals of 3.67 × 10⁻⁵ s each.

2.6. Design and Development of the Test Bench for Validating the Optimized Fan

For experimental validation of the optimized fan, a controlled experimental test bench was designed and developed. The test bench consisted of two main sections: a rotating part (fan) and a stationary part, as shown in Figure 11a,b, respectively. The main components of the test bench were a fan volute having a width of 804 mm, interchangeable fan blades, adjustable airflow guide plates, a variable-speed electric motor, transparent acrylic walls, pulleys and belts for power transmission, a rotating shaft, and a supporting structural frame equipped with wheels for mobility and perforated plates. This setup was used to validate the new fan and conduct a comparative analysis with the existing fan.

2.6.1. Airflow Velocity Measurement at the Outlets of the New and Existing Fans

The airflow velocity at the fan outlets (Duct 1, Duct 2, and Duct 3) of both optimized and existing fan was measured at six points along the outlet width (115, 230, 345, 460, 575, and 690 mm as designated as A, B, C, D, E and F respectively) using a VT100 model anemometer (accuracy: ±0.05 m/s) from KIMO, Paris, France, as shown in Figure 12a,b and Figure 13a,b. The DoE for experimental validation is detailed in

Table 1. DoE for comparative analysis under test bench conditions.

Test No.	GPA (No. 1) (°)	GPA (No. 2) (°)	Fan Speed (rpm)
1	26	32	1000
2	38	32	1000
3	26	32	1200
4	38	32	1200
5	32	26	1000
6	32	38	1000
7	32	26	1200
8	32	38	1200
9	26	26	1100
10	38	26	1100
11	26	38	1100
12	38	38	1100

. To ensure accuracy, each measurement was repeated three times under the same conditions. The coefficient of variation (CV) was calculated using Equation (13). According to ASAE [38], a CV below 20% is generally considered acceptable.

$$CV={\displaystyle \frac{\sigma }{\mu }}\times 100\%$$

(13)

where σ is the standard deviation, m/s; μ is the mean of the airflow velocity, m/s.

2.6.2. Airflow Distribution with Perforated Plate and Validation of the CFD Simulation for the Optimized Fan

To assess the airflow spatial variability, six data collection points were distributed lengthwise along the sieve at 150, 350, 550, 750, 950, and 1150 mm. Additionally, six positions were selected across the sieve’s width direction at 65, 195, 325, 455, 585, and 715 mm. The airflow velocity distribution was created above positive (+ve) 50 mm and below negative (−ve) 50 mm on the perforated plate, as shown in Figure 14a,b.

The fan’s performance was compared against the simulated data to assess the CFD model accuracy. The coefficient of determination (R²) was calculated to quantify the agreement between the measured and simulated data using the standard sum of squares formula. It was calculated using Equation (14):

$$R^2={\displaystyle \frac{{\sum (y_i-{\widehat{y}}_i)}^2}{{\sum (y_i-{\overline{y}}_i)}^2}}$$

(14)

where $y_i$ is the measured airflow velocity values, m/s; ${\widehat{y}}_i$ is the simulated airflow velocity values, m/s; and ${\overline{y}}_i$ is the mean of the measured airflow velocity values, m/s.

The absolute error, which represents the difference between the measured and simulated values, was used as an indicator of the CFD model’s predictive accuracy. It was calculated using Equation (15):

$${\epsilon }_{abs}= \left|V_{MEA}-V_{CFD}\right|$$

(15)

where ${\epsilon }_{abs}$ is the absolute error, m/s; $V_{MEA}$ is the measured airflow velocity value, m/s; $V_{CFD}$ is the simulated airflow velocity value, m/s.

3. Results and Discussion

3.1. Total Static Pressure Reduction (P)

Figure 15a illustrates that the threshed material accumulated most prominently in the front region of the sieve, where the layer height reached a peak of 0.22 m in Zone 3. In comparison, both the middle and rear sieve regions displayed significantly reduced material thickness. The computed fluidized grain layer height, as derived from Equation (2), reached a maximum of 0.06 m in the same front sieve region, as shown in Figure 15b. Grain distribution along the sieves was uneven: 54% in the front, 34% in the middle, and 12% in the rear, respectively., as detailed in Figure 15c. A detailed breakdown of the threshed material composition, consisting of full grain, blighted grain, grains with long shoots, grains with short shoots, short straw, and chaff, is presented in Figure 16. From Figure 16a, the average composition across the sieve was dominated by full grain (74.22%), followed by chaff (14.68%), and smaller proportions of other components. When segmented by sieve zones as shown in Figure 16b, the front sieve region (0–450 mm) had a concentration of full grain (27.28%) and chaff (2.88%). The middle sieve region (450–1050 mm) presented 25.58% full grain and 4.59% chaff, whereas the rear sieve region (1050–1500 mm) contained 21.36% full grain but had the highest proportion of chaff at 7.21%. Furthermore, the spatial distribution of the fluidized grain layer exhibited a declining pattern from front to rear: 36.76% of the fluidized grain was found in the front, 34.46% in the middle, and only 28.78% in the rear sieve region.

The static pressure reduction (P) was calculated using Equation (1) and detailed in

Table 2. Total static pressure reduction (P) and airflow drag coefficients ($c_f$) at nine zones.

	No. of Zone
Items	Zone1	Zone 2	Zone 3	Zone 4	Zone 5	Zone 6	Zone 7	Zone 8	Zone 9
Pressure reduction, in Pa	396.8	406	522	222.3	244	448.2	150	153.8	183.8
Drag coefficient	6.41	6.67	8.49	2.54	2.81	5.06	1.24	1.30	1.57

. The maximum static pressure reduction (P) of 522 Pa was used in the design process. The airflow velocity of the fan outlets was designed to be 24 m/s to meet the rice combine harvester feed rate of 7 kg/s. The dynamic pressure was calculated and resulted in a value of q = 383 Pa. The drag from the sieves was 1.04 [24] and contributed to a pressure reduction of 398.32 Pa. Therefore, considering the pressure required to remove straw and chaff, the drag from the cleaning sieves, and the inherent static pressure of the fan (assumed to be 25% of the internal drag), the total static pressure requirement for the fan was calculated to be 1150.32 Pa. This value was applied in the blade width calculation to ensure the fan generates enough pressure to overcome the airflow drag. The drag coefficients due to the fluidized grain (c_f) were calculated using Equation (2) [32].

3.2. Design Analysis of the New Fan Blade

Based on the parameters β = 15%, μ = 0.26; $\rho$ = 1.225 kg/m³; w_c = 2.0 m; v = 1.1 m/s; γ = 2.1; and p = 1.02 kg/m², the theoretical airflow rate (Q_v) was determined to be 3.3 m³/s using Equation (4). Substituting all parameters into Equation (5), the calculated specific speed (Ns) was found to be 11.1. This value indicates that the fan is of the centrifugal type, and the characteristics of the centrifugal fan are detailed in

Table 3. Characteristics of centrifugal fan blades.

Fan Type	b/D	Maximum Efficiency (%)	Number of Blades
Radial blade	0.35–0.45	60–70	6–8
Backward curved	0.25–0.45	75–90	8–12
Forward curved	0.50–0.60	55–60	16–20

b—blade width, mm; D—impeller diameter, mm.

.

The blade shape was selected based on the operational requirements of the cleaning system, including the need to provide uniform airflow across the sieve regions, withstand variations in cleaning load, and generate the required airflow velocity for high-throughput rice harvesting. For the preliminary design, the most appropriate fan blade type was selected to meet the performance requirements of the cleaning system. Accordingly, for the structural optimization of the fan blade design, a backward-curved blade, as shown in Figure 17, with the exit blade angle (β₂ < 90°) was selected due to its higher efficiency compared to forward-curved blades, offering a maximum efficiency of 75–90%. The number of blades was set to n = 12, as detailed in Table 3. Blower-type cleaning fans in combine harvesters are typically designed as double-suction centrifugal fans to ensure high airflow capacity and improved pressure stability. Therefore, the L9-3 × 12 fan type, characterized by its double suction and backward-curved blades, was selected, as shown in Figure 17a,b from the fan design manual [33], as the reference model for redesign.

Design Analysis of the Impeller Inlet Diameter and Blade Width

The impeller inlet diameter and blade width were designed based on the required airflow rate (Q_v = 3.3 m³/s) and pressure reduction (P = 1150.32 Pa) to achieve the necessary airflow velocity for effective grain cleaning. The pressure coefficient and flow coefficient were calculated using Equations (9) and (10), respectively, resulting in ψ = 3.0 and ϕ = 0.8. The calculated pressure coefficient indicates that the proposed fan blade is capable of generating a relatively high static pressure to overcome the system drag. The tangential velocity u₂ (m/s) at the exit of the impeller is calculated using a fan speed of 1200 rpm and an impeller outer diameter of D₂ = 0.395 m. Substituting these values into Equation (8) yields a tangential velocity of u₂ = 24 m/s. Substituting all the relevant parameters and coefficients into Equation (9), the impeller inlet diameter was calculated to be D₁ = 0.380 m. The blade width was calculated as b′ = 0.268 m using Equation (10), and the impeller width (w) was determined as w = (1.15–3.15) b′, resulting in w = 0.80 m. A summary of the design optimization parameters is presented in

Table 4. Summary of the design parameters for fan geometry.

Parameters	Symbol	Unit	Model 1	Model 2	Model 3
Hub diameter	D0	mm	50	50	50
Impeller inner diameter	D1	mm	380	380	380
Impeller outer diameter	D2	mm	395	395	395
Blade spacing at the bottom	t1	mm	21	21	21
Blade spacing at the top	t	mm	48	31	14
Blade thickness	s	mm	3	3	3
Blade inlet angle	$\beta$1	°	92.64	92.64	92.64
Blade outlet angle	$\beta$2	°	62.18	62.18	62.18
Blade curvature 1	BC1	mm	147	147	147
Blade curvature 2	BC2	mm	97.62	97.62	97.62
Blade inlet curvature	BIC	mm	135	135	135
Impeller width	$w$	mm	800	800	800
Blade clocking angle	$\delta$	°	0	5.5	10.5

.

3.3. CFD Simulation Results for Selecting the Optimal Fan Configuration

From Figure 18a, for the fan model with a clocking angle of 0°, the airflow velocity at Duct 1 was 8.45 m/s; at Duct 2, it increased to 10.04 m/s; and at Duct 3, it slightly increased to 12.11 m/s. The airflow rate was 1.13 m³/s at Duct 3, 1.04 m³/s at Duct 1, and 0.82 m³/s at Duct 2. The total rate from the three ducts was 2.99 m³/s, which is lower than the design rate. In contrast, for the fan model with 5.5° clocking angle, as shown in Figure 18b, the airflow velocity at Duct 1 was 9.48 m/s, at Duct 2 it decreased to 11.81 m/s, and at Duct 3, it increased significantly to 13.29 m/s. The airflow rate was 1.20 m³/s at Duct 3, the lowest rate was 0.90 m³/s at Duct 2, and it was 1.02 m³/s at Duct 1. The total airflow rate from the three ducts was 3.12 m³/s, which is also lower than the required airflow rate. For the fan model with a 10.5° clocking angle, as shown in Figure 18c, the velocity at Duct 1 was 10.55 m/s, while Duct 2 and Duct 3 recorded 12.85 m/s and 14.95 m/s, respectively. The airflow rate at Duct 1 was 1.12 m³/s, Duct 2 showed 0.92 m³/s, and Duct 3 was 1.42 m³/s, resulting in a total airflow rate of 3.46 m³/s that meets the required airflow. The relatively high velocity at Duct 1 is particularly beneficial, as it corresponds to the front region of the cleaning sieve, where initial grain separation and fluidization begin. Therefore, the 10.5° configuration makes it the most effective setup among the simulated fan models. The red line indicates the airflow distribution at the volute tongue near duct 1. The red line indicates the airflow distribution at the volute tongue near duct 1, with the 10.5° clocking angle showing the highest airflow distribution.

From Figure 19a, it can be seen that the fan model with a 0° clocking angle generated the highest total pressures across the ducts, with 335.78 Pa at Duct 1, 385.60 Pa at Duct 2, and 355.45 Pa at Duct 3. The 5.5° clocking angle fan showed lower pressures, recording 360.33 Pa at Duct 1, 296.75 Pa at Duct 2, and 361.55 Pa at Duct 3, suggesting uneven pressure distribution. For the 10.5° clocking angle, the pressures were more balanced: 395.24 Pa, 365.69 Pa, and 415.25 Pa at Ducts 1, 2, and 3, respectively. This resulted in a total static pressure of 1176.18 Pa, which meets the required design specification. The red line indicates the pressure distribution, with the 10.5° clocking angle showing the highest pressure distribution.

The simulation results revealed that the fan models with 0° and 5.5° blade clocking angles provided slightly improved airflow uniformity compared to the baseline but still exhibited noticeable unevenness between the front and rear zones of the cleaning sieve, as shown in Figure 20a,b. In contrast, the fan with a 10.5° clocking angle demonstrated the most uniform and balanced airflow distribution across the entire sieve surface, as shown in Figure 20c. Additionally, the simulated airflow velocity in Duct 1 was higher for the 10.5° clocking configuration than for the 0° and 5.5° setups, indicating improved performance in directing airflow. Based on these results, the 10.5° clocking angle fan was selected for further experimental validation. It achieved superior airflow velocity above the sieve, ranging from 8.5 to 6.2 m/s in the front region, 6.0 to 5.0 m/s in the middle region, and 4.0 to 3.5 m/s in the rear region, highlighting its effectiveness in delivering consistent airflow across the cleaning area. The red line indicates the airflow distribution inside the cleaning system, with a 10.5° clocking angle showing better airflow distribution.

3.3.1. Effects of Blade Clocking Angle, Blade Spacing, and Fan Speed on Fan Performance

At Duct 1, as detailed in

Table 5. Response surface analysis for airflow velocity at Duct 1.

Source	LogWorth		p-Value
$\delta$ (0, 5.5, and 10.5)	11.693		0.00010
t (15, 31, and 48)	2.669		0.00214
δ × t	2.177		0.00665
N (1000, 1100, and 1200)	2.174		0.00670
δ × δ	0.555		0.27831
t × t	0.552		0.28031
N × N	0.094		0.80505
δ × N	0.077		0.83792
t × N	0.038		0.91652

δ—blade clocking angle in ^o; t—blade spacing in mm; N—fan speed in rpm.

, the blade clocking angle was identified as the most dominant factor, with a LogWorth value of 11.693 and a p-value of 0.00010, indicating a highly significant effect. This result confirms that increasing the clocking angle, particularly from 0° to 10.5°, substantially improves airflow velocity in this region. Additionally, blade spacing (LogWorth = 2.669, p = 0.00214) and fan speed (LogWorth = 2.174, p = 0.00670) had statistically significant effects.

For Duct 2, as detailed in

Table 6. Response surface analysis for airflow velocity Duct 2.

Source	LogWorth		p-Value
$\delta$ (0, 5.5, and 10.5)	11.120		0.00010
t (15, 31, and 48)	10.049		0.00011
N (1000, 1100, and 1200)	3.363		0.00043
δ × t	1.951		0.01120
δ × δ	0.941		0.11461
t × t	0.479		0.33159
δ × N	0.424		0.37647
t × N	0.228		0.59216
N × N	0.100		0.79460

δ—blade clocking angle in ^o; t—blade spacing in mm; N—fan speed in rpm.

, both the blade clocking angle and blade spacing were found to be highly significant. The clocking angle showed a LogWorth of 11.120 (p = 0.00010), while blade spacing had a LogWorth of 10.049 (p = 0.00011), indicating that changes in both parameters strongly affect airflow performance. Increasing the clocking angle to 10.5° and enlarging the spacing between blades led to improved airflow velocity in this middle region. The fan speed also played a meaningful role (LogWorth = 3.363, p = 0.00043), enhancing the airflow through this duct.

At Duct 3, as detailed in

Table 7. Response surface analysis for airflow velocity Duct 3.

Source	LogWorth		p-Value
$\delta$ (0, 5.5, and 10.5)	12.943		0.00005
t (15, 31, and 48)	12.269		0.00010
δ × δ	7.710		0.00015
N (1000, 1100, and 1200)	5.032		0.00021
$\delta$ × t	2.933		0.00117
t × N	0.748		0.17854
$\delta$ × N	0.645		0.22633
N × N	0.373		0.42369
t × t	0.229		0.58963

δ—blade clocking angle in ^o; t—blade spacing in mm; N—fan speed in rpm.

, the blade clocking angle showed as the most significant factor, with a remarkably high LogWorth of 12.943 and a p-value < 0.00001. Blade spacing showed a comparable effect, with a LogWorth of 12.269 and the same high level of significance (p < 0.00001). These results confirm that geometric modifications to blade orientation and spacing are the key drivers of airflow behavior in this region. Furthermore, fan speed also had a statistically significant effect (LogWorth = 5.032, p = 0.00001). The RSM results indicated that a blade clocking angle of 10.5°, a blade spacing of 15 mm, and a fan speed of 1200 rpm significantly enhanced the airflow to 10.55 m/s at Duct 1, 12.85 m/s at Duct 2, and 18.05 m/s at Duct 3. Therefore, this fan configuration was selected.

3.3.2. Fan Performance Curve Under Load Conditions

The performance curves of three fan models were generated under simulated load conditions. These conditions represent the effects of airflow drag on airflow rate of the fan outlet. From Figure 21a–c, a steeper slope in the performance curve was observed for Duct 1, Duct 2, and Duct 3 when operating with the 10.5° clocking angle configuration. This steeper gradient indicates that the 10.5° fan model is capable of maintaining higher airflow rates. However, in the cases of clocking angles 0° and 5.5°, the load coefficient–rate curves are relatively flat, indicating a decrease in airflow rate as the load increases, as shown in Figure 21a–c.

The CFD simulation results for the model with a 10.5° blade clocking angle revealed that, with GPAs set to No. 1 = 32° and No. 2 = 32° and a fan speed of 1200 rpm, optimal airflow performance was achieved. The airflow velocity ranged from 8.55 to 10.75 m/s in the front region, 7.15 to 8.45 m/s in the middle region, and 5.35 to 6.75 m/s in the rear region. Subsequently, experimental validation was performed on the test bench to verify the CFD model for the fan with a 10.5° clocking angle.

3.4. Analysis of Airflow Distribution at the Outlets of the New Fan

The test condition with GPAs No. 1 = 32° and No. 2 = 32° and a fan speed of 1200 rpm was identified as optimal. Under this condition, the maximum airflow velocity at Duct 1 was 10.20 m/s as shown in Figure 22a, indicating improved airflow supply to the front sieve region, which is critical for fluidizing heavier particles. Similarly, Duct 2 recorded a velocity of 13.07 m/s as shown in Figure 22b, while Duct 3 showed the highest velocity of 14.85 m/s as shown in Figure 22c. These results suggest that the optimized fan ensures more evenly distributed airflow, which improves the cleaning performance. The volumetric airflow rate delivered through the fan outlets under various operating conditions was systematically assessed to evaluate the fan’s performance. As shown in Figure 22c, airflow was measured across eight different test conditions. Among these, No. 1 = 32° and No. 2 = 32°, with a fan speed of 1200 rpm, achieved a total airflow rate of approximately 3.3 m³/s. Consequently, the combined airflow from all three outlet ducts met the specified airflow requirements of the cleaning system. This rate ensures effective grain fluidization over the sieve surface, improving the separation between grain and chaff and increasing overall cleaning efficiency.

The comparative analysis of airflow velocity at the fan outlets between the existing and optimized fan designs revealed that the optimized fan significantly improved airflow performance at Duct 1 and Duct 2 of the cleaning system while slightly reducing airflow at Duct 3. Specifically, the test condition with GPAs No. 1 = 32° and No. 2 = 32° and a fan speed of 1200 rpm was identified as the optimal setting among all tested configurations. Under this condition, the maximum airflow velocity at Duct 1 increased from 6.93 m/s with the existing fan to 10.20 m/s with the optimized fan, indicating enhanced airflow delivery to the front sieve region for fluidizing heavier material, as shown in Figure 23a. Similarly, at Duct 2, airflow velocity improved from 11.7 m/s with the existing fan to 13.7 m/s with the optimized fan, as shown in Figure 23b. However, at Duct 3, the new fan showed a slightly lower velocity of 14.85 m/s compared to 17.17 m/s with the existing fan, as shown in Figure 23c.

The test condition with GPAs No. 1 = 32° and No. 2 = 32° and a fan speed of 1200 rpm was identified as the optimal setting among all tested configurations. Under these optimal conditions, the CV calculated using Equation (13) for the optimized fan was 15.83% for Duct 1, 13.68% for Duct 2, and 14.68% for Duct 3—all within the acceptable range (CV < 20%). In contrast, under the same conditions, the existing fan exhibited significantly higher variability, particularly in Ducts 1 and 2, with CV values of 47.22% and 28.29%, respectively, while airflow at Duct 3 was relatively uniform, with a CV of 7.52%, as detailed in

Table 8. Comparative analysis of airflow uniformity at the new and existing fan outlets.

	Existing Fan, CV (%)			Optimized Fan CV (%)
Conditions	Duct 1	Duct 2	Duct 3	Duct 1	Duct 2	Duct 3
1	54.78	11.39	17.63	4.3	3.33	12.42
2	18.39	1.56	9.49	18.78	13.57	6.14
3	21.07	23.91	6.86	19.55	14.02	9.39
4	33.94	26.27	8.87	19.22	3.79	20.78
5	28.69	28.06	6.20	15.97	12.68	13.48
6	26.36	43.26	11.99	13.53	10.31	16.64
7	30.69	33.22	4.19	4.22	7.24	6.36
8	37.18	25.85	19.26	7.73	4.50	5.24
9	38.27	32.61	11.52	5.67	5.63	7.91
10	47.22	28.29	7.52	15.83	13.68	14.68
11	39.09	21.09	10.49	19.66	7.46	7.80
12	32.5	22.05	11.40	12.02	8.25	7.75

. This comparison indicates that the optimized fan provides more uniform airflow distribution. Compared with the original fan, the reductions in CV values were 66%, 51.6%, and 27.5% for Ducts 1, 2, and 3, respectively, indicating a substantial improvement in airflow uniformity across all outlets.

3.5. Analysis of Airflow Velocity of the New Fan Using a Perforated Plate

Detailed analysis was conducted under working load conditions using the optimized fan. Accordingly, from Figure 24a, it can be seen that under the test condition with GPAs set at No. 1= 32° and No. 2 = 32° and a fan speed of 1200 rpm, the airflow velocity distribution below the perforated plate showed an optimal condition. The airflow velocity in the front region reached 8.55–10.75 m/s, while the middle regionrecorded 6.45—7.15 m/s, and the rear region measured 4.75–3.85 m/s. These velocity values are considered sufficient to ensure that adequate airflow passes through the sieve to effectively fluidize the grain layer. As shown in Figure 24b, under this test condition, the airflow velocity above the perforated plate was measured to be in the range of 0.55–2.95 m/s in the front region, 0.55–1.35 m/s in the middle region, and 0.75–2.25 m/s in the rear region.

Validation of the measured airflow velocities and CFD simulation results, as shown in Figure 25, indicates that the numerical model provides a reliable representation of the actual airflow behavior within the cleaning system. A linear regression analysis between the measured and CFD-predicted airflow velocities yielded the regression equation CFD = 0.815 × Measured + 2.033, with a coefficient of determination R² = 0.927, calculated using Equation (14). This high R² value reflects a strong correlation between simulated and experimental data. The maximum absolute error was calculated using Equation (15), and the observed value was 0.35 m/s, which refers to the highest difference between the measured and simulated airflow velocities at corresponding points, indicating the peak deviation in validation. And all percentage errors are below 5%, which confirms that the CFD model accurately captures the airflow distribution characteristics under the influence of cleaning drag.

3.6. Cost–Benefit Analysis of the Optimized Fan

The cleaning fan developed in this study can be applied to combine harvesters and is suitable for harvesting various crops such as rice, wheat, and soybeans. Forecast results indicate that the fan can reduce grain cleaning loss by approximately 20%. Based on each combine harvester operating over 15 hectares per year with an average yield of 9000 kg/hectare, it is estimated that each harvester can reduce grain loss by about 700 kg annually. At an average grain purchase price of CNY 2.5/kg, this translates to an annual income increase of approximately CNY 1750 per machine. The equipment cost can thus be recovered within the first year, and with an estimated service life of 8 years, the cleaning fan offers significant economic benefits.

4. Conclusions

• This study presents a novel multi-blade cleaning fan design that significantly improved airflow uniformity and distribution at the fan outlet, thereby enhancing the cleaning efficiency of longitudinal-flow rice combine harvesters. Through comprehensive CFD analysis and experimental validation, the newly designed fan demonstrated enhanced airflow uniformity, contributing to more stable and efficient cleaning performance. Additionally, it is affordable and well-suited for mass production;
• A new cleaning fan blade was designed based on the required static pressure (P) and the desired airflow rate (Q_v) for the cleaning load. Using a maximum static pressure of 522 Pa, the theoretical airflow rate was calculated as 3.3 m³/s. Three fan models were developed with blade clocking angles of 0°, 5.5°, and 10.5°, designed according to these parameters and standard fan design principles. The fan features a 10.5° clocking angle, achieving airflow velocities of 10.55, 12.85, and 14.95 m/s at Ducts 1, 2, and 3, with corresponding rates of 1.12, 0.92, and 1.42 m³/s, respectively. The total airflow rates was 3.46 m³/s, which meets the design requirements;
• A test bench was developed to evaluate the performance of the newly designed fan and compare it with the existing four-bladed fan. The airflow velocity distribution at the fan outlets was analyzed based on experimental measurements. Among the tested configurations, the setup with GPAs No. 1 = 32° and No. 2 = 32° and a fan speed of 1200 rpm was identified as optimal. Under these conditions, the optimized fan showed good uniformity, with CV of 15.83%, 13.68%, and 14.68% at Ducts 1, 2, and 3, respectively—all within the acceptable limit (CV < 20%). In contrast, the existing fan exhibited much higher variability under the same conditions, with CVs of 47.22% at Duct 1 and 28.29% at Duct 2, while Duct 3 showed relatively uniform airflow (CV = 7.52%).

Therefore, in comparison to the conventional centrifugal fan with four blades, the optimized multi-blade design demonstrates superior aerodynamic performance and is more appropriately aligned with the operational requirements of modern rice varieties.