Mechanistic analysis of fracture-controlled canola dehulling via finite element modeling

Abstract

Canola is an important oilseed crop valued for its oil and protein, both of which reside primarily in the kernel. Dehulling separate the hull from the kernel and is important for improving product quality. This study aimed to engineer a controlled fracture-propagation pathway that maximizes hull removal while preserving kernel integrity. Canola was first hydrated and dried using a fluidized bed dryer (FBD) to tailor the mechanical properties, followed by mechanical loading using a roller machine that generated either compression-dominated or shear-dominated stresses. Finite element modeling and experimental validation were combined to reveal how pretreatment and stress mode shape fracture initiation, propagation, and hull detachment dynamics. The 2D model characterized the mechanical properties of the hull and kernel based on uniaxial compression tests, while the 3D model simulated deformation and stress distribution during the dehulling process. High-speed imaging captured real-time dehulling behavior, and synchrotron-based X-ray tomography visualized internal fracture patterns and kernel preservation.

FBD pretreatment significantly reduced hull strength and increased kernel elasticity, specifically the lowest ultimate stress of 11 MPa in the hull and the elasticity of 30.5 MPa with the kernel, creating a favorable mechanical contrast that promoted preferential hull rupture and clean hull–kernel separation while minimizing kernel damage. The shear-dominant mode further promoted directional crack propagation and hull detachment, resulting in substantially improved recovery of intact, hull-free kernels. These findings establish a mechanistic foundation for precision-controlled fracture and separation processes across food, agricultural, and biomaterial systems.

Graphical abstract

Highlights

• •Engineered a controlled fracture pathway enabling selective canola dehulling.
• •Fluidized-bed drying tailored hull–kernel mechanical properties, enabling selective dehulling.
• •Shear-dominant loading guided fracture along the hull–kernel interface, preserving kernel integrity.
• •Integrated experiments, 3D FE simulation, and synchrotron X-ray imaging to visualize fracture dynamics.
• •Established a mechanistic and materials-based foundation for precision, energy-efficient oilseed dehulling.

1. Introduction

Canola (Brassica napus subsp. napus) is one of the world's most valuable oilseed crops, prized for its high oil and protein content (Xie et al., 2019). In Canada alone, canola is a major economic crop with production in 2023/2024 of 18.8 million tonnes (MT) (FAO, 2024). However, the presence of fibrous hulls significantly limits the efficiency of oil extraction and the nutritional quality of canola meal. While the meal can be incorporated into animal diets as a protein source, it can influence feed palatability, intake, and the production of milk, meat, or eggs (Raboanatahiry et al., 2021; Zhu et al., 2019). Previous studies have reported that the hull fraction remaining in the meal contains anti-nutritive compounds, such as tannins and crude fiber, which reduce the palatability and the high fiber in pig diets negatively affect nutrient digestibility and energy value (Clandinin, 1961; Lyu et al., 2018). Dehulled rapeseed meal has been identified as a superior alternative, improving phosphorus digestibility without altering body weight in pigs (Mejicanos et al., 2018).

However, current dehulling technologies, such as flaking, centrifuging, and abrasing, apply mechanical force to break the seed into irregular fragment and fines, causing substantial structural damage (Carré et al., 2016; Ikebudu et al., 2000; Carré and Loison, 2021). These uncontrolled fracture behaviors generate particles with diverse sizes and shapes, which not only reduce the efficiency of subsequent separation processes (Greffeuille et al., 2006; Heinze et al., 2016) but also cause significant oil and protein losses, as protein concentrated within the kernel is transferred to the hull fraction during separation. Most previous studies have relied on empirical, trial-and-error optimization of dehulling conditions, focusing on performance outcomes rather than understanding the underlying fracture mechanics. As a result, seed fracture remains largely uncontrolled, leading to inconsistent separation efficiency and unnecessary kernel damage. To overcome these limitations, the present study establishes a mechanistic framework for engineering a dehulling process that controls fracture propagation to maximize hull removal while preserving kernel integrity. This approach can be readily adapted to existing industrial operations and enhances downstream separation by exploiting the difference in aerodynamic terminal velocity between the hull and kernel for efficient air classification. Since fracture pathways are not random but depend strongly on factors such as grain hardness, internal microstructure, moisture content, and the type and direction of mechanical loading, controlling these parameters is essential for achieving efficient and selective dehulling. Ideally, the fracture occurs along the hull while leaving the internal kernel intact. Given the structural complexity of canola seeds, the desired mechanical action should promote fracture propagation along the hull membrane, thereby loosening or separating it from the underlying kernel without causing internal penetration or damage.

To achieve such selective fracture behavior, recent studies suggest that pre-treating canola seeds by moistening and heating with a fluidized bed dryer (FBD) facilitates hull–kernel separation by creating an interfacial air gap that serves as a preferential fracture pathway. This pathway weakens the adhesion between the hull and kernel while guiding crack propagation along the hull, enabling the hull to rupture while the kernel remains intact. Moisture content appears to be the dominant factor influencing dehulling performance, as it modifies the mechanical properties of both the hull and the kernel; however, detailed mechanistic insights remain limited (Yin et al., 2025). Therefore, tailoring seed moisture and mechanical properties to enhance structural robustness against applied forces —allowing the hull to crack without damaging the kernel—and applying directional mechanical loading to steer fracture propagation represent promising strategies for realizing the ideal dehulling outcome.

Although the mechanical behavior of canola seeds at different moisture contents has been studied (Davison et al., 1975; Cenkowski et al., 1992; Izli et al., 2009), most experiments used simple rewetting to reach target moisture levels, thereby overlooking how different pretreatment methods may alter seed mechanical properties even at the same moisture content. Results showed that as moisture content increased, the ultimate fracture force decreased, and the compression behavior could be described by an elastoplastic model (Yuan et al., 2022). Martinez-Soberanes et al. (2022) further compared the mechanical responses of canola seeds under shear and compression loading, reporting that at the same moisture content, the ultimate strength under shear was approximately four times lower than under compression. Morphological observations after hull cracking also revealed that seeds deformed less under shear stress than under compression: at a moisture content of 5.8 % (w.b.), the kernel retained its shape when the hull fractured, whereas higher moisture levels led to varying degrees of plastic deformation within the kernel. Despite these valuable insights, most prior studies have focused primarily on quantifying breaking forces rather than elucidating the underlying fracture mechanisms or the influence of pretreatment on mechanical response. Furthermore, the frictional behavior of canola seeds on various material surfaces has also been examined (Izli et al., 2009; Çalışır et al., 2005; Xu et al., 2019). These studies showed that the static friction coefficient increases with moisture content, with the highest values observed on rubber, followed by plywood, galvanized iron, glass, aluminum, and stainless steel (Izli et al., 2009). While these findings highlight the role of moisture in seed–surface interaction, the potential influence of pretreatment methods, which can modify surface texture and structural stiffness, have been largely overlooked. To address this gap, the present study systematically investigates the mechanical behavior of canola seeds following controlled pretreatment using a fluidized bed drying (FBD) system.

Dehulling technology also plays a crucial role in determining separation performance, as it provides the directional mechanical loading needed to steer fracture propagation and orient cracks along desired pathways. To achieve such control, it is essential to understand the interaction between seeds and dehulling equipment under different operating conditions, particularly the mechanisms of deformation and stress distribution. However, the complex cellular structure of canola seeds makes it difficult to directly measure internal stresses during mechanical loading. As a result, the finite element method (FEM) has become a valuable tool for modelling objects, predicting stresses, and analyzing mechanical behavior (Hernández and Bellés, 2007; Khodabakhshian and Emadi, 2015; Hasseldine et al., 2017). Martinez-Soberanes et al. (2022) developed a compression FEM model for canola seeds, determining the modulus of elasticity of the kernel and hull at 5.8 % (w.b.) as 490 MPa and 1180 MPa, respectively. However, this model did not accurately represent practical dehulling conditions, as it considered only compression stress, whereas industrial dehulling involves combined compression–shear loading. To address these limitations, the present study integrates finite element modeling with roller-based dehulling experiments to establish a mechanistic understanding of canola fracture and separation behavior. The FEM model was developed to simulate the dehulling process performed on a roller machine capable of operating in two distinct modes: (1) a compression mode, in which the rollers rotate in opposite directions to generate primarily normal forces that compress and squeeze the hull, and (2) a shear mode, in which both rollers rotate in the same direction to create combined normal and shear forces that promote sliding and tearing. These two loading conditions reproduce the dominant stress states observed in industrial dehulling and enable qualitative validation of FEM-predicted deformation and fracture patterns against experimental observations. Corresponding dehulling experiments were conducted using the same construction roller system, with high-speed imaging employed to capture fracture propagation and kernel integrity during each loading mode.

Overall, this study establishes a mechanistic framework linking material property, stress transfer behavior, and fracture propagation to optimize canola dehulling. A combination of experimental testing, numerical modeling, and advanced imaging was employed to reveal the mechanisms underlying the engineered fracture pathway and resulting dehulling performance. The mechanical properties of FBD-pretreated canola (hulls and kernels respectively) were characterized through uniaxial compression testing, elastoplastic model and 2D FEM analysis. A 3D FEM model was further developed to illustrate stress distribution and deformation under both compression and shear loading modes in the roller machine. Complementary dehulling experiments were performed, with high-speed imaging capturing real-time fracture dynamics and synchrotron-based X-ray micro-computed tomography (μCT) visualizing internal structural after dehulling. Together, these integrated approaches bridge the gap between empirical dehulling studies and mechanistic understanding, providing critical insight into fracture-controlled dehulling and offers a scientific basis for developing efficient, kernel-preserving industrial dehulling technologies.

2. Methods and materials

2.1. Seed material and pretreatment

InVigor™ hybrid canola seed (B. napus) was harvested in 2021 from Colborn farm, SK, Canada. Initial moisture content was 5.4 % wet mass basis (w.b.). Seeds were prepared by moistening at 75 $°C$ for 30 min then storing at −4 $°C$ for 24 h in the plastic resealable bag, water was dispersed evenly inside the seed, and the moisture content reached 46 % (w.b.) waiting further process for specific experiments. Moisture content was determined in triplicate following the ASAE Standard S352.2 (ASAE, 1998).

2.2. Mechanical testing

2.2.1. Compression test setup

The compression tests were conducted at room temperature using a texture analyzer (TA.XT_plus, Stable Micro Systems, Godalming, UK) with 5.00 Kg cell load and 0.05 mm/s loading rate. The displacement and load force were recorded which representing the deformation and force of the seed before rupture.

2.2.2. Seed moisture content effects on mechanical properties

Canola was pretreated to achieve different moisture contents, and their mechanical properties were measured. Canola (350 g) was pretreated by a fluidized bed dryer (Lab-line Instruments, Inc. Melrose Park, IL, USA). Based on the result from the pre-experiment, the drying conditions was design as the heating temperature of 50 $°C$, an air flow rate of 6.5 m/s, and heating cycle of 3, 5, 6, 7 and 9 repetitions. Each heating cycle consisted of 10 min of heating followed by 5 min of cooling at room temperature. The resulting moisture contents of the seed was 26 %, 16 %, 12 %, 8.8 % and 5.1 % (w.b.), respectively. For each moisture content, 15 intact canola seeds with a diameter of 2 mm were selected and tested. The force–displacement (F–D) curve for each moisture represents a fitted curve based on the results from the 15 individual seeds.

2.2.3. Mechanical properties of the kernel

To characterize the behavior of the kernel under the mechanical stress, kernels were obtained from seeds with moisture contents of 8.8 %, 16 %, and 26 % (w.b.). For each moisture content, 15 kernels were obtained by manually removing the hull from seeds with a diameter of 2 mm and were subsequently subjected to compression testing. The representative force–displacement (F–D) curve for each moisture content was generated by fitting the data obtained from the 15 individual kernels.

2.2.3.1. Kernel orientation for compression test

Without the protective hull, the kernel becomes more vulnerable and exhibits inconsistent behavior during compression testing. These inconsistencies arise from variations in the orientation of kernel placed under the compression load. To ensure consistent and reliable results, the orientation of the kernel during compression testing was standardized.

Synchrotron-based X-ray imaging, which is a non-destructive method, was applied to visualize the internal anatomy of the canola seed. When the seed was sectioned along the ridge (Fig. 1 A), it revealed a kernel structure consisting of a radicle and two cotyledons (Fig. 1 B). The radicle is bent and nestled between the cotyledons (Fig. 1 C-1), with its tip splitting into two parts that gradually develop into the cotyledons located on either side (Figs. 1 C–2). Based on this developmental pattern, a coordinate system was established for the kernel. The point where the radicle splits was defined as the origin (0,0). The cotyledon on the right side of the split was designated as the right cotyledon, extending in the positive X-direction, while the left cotyledon extends in the negative X-direction. The positive Y-direction points downward, toward the ground. This coordinate system not only standardizes the positioning of the kernel during the compression test but also provides potential insights into seed germination and metabolic processes. During the compression test (Fig. 1 D), the junction of the radicle and cotyledons was consistently aligned under the center of the loading cell, with the split end of the radicle always facing outward.

Fig. 1.

Anatomy of canola. (A) 3D reconstruction of a whole canola seed. (B) Internal structure of the seed. R: radicle; C: cotyledon. (C) Cross-section of different cutting planes. LC: left cotyledon; RC: right cotyledon. (D) Kernel orientation during compression test.

Fig. 2.

Dehulling models. (A) 2D dehulling model of compression test, consisted of hull and kernel. (B) 3D dehulling model of the roller machine containing a canola seed, where the hull was simplified as an annular layer surrounding the kernel.

2.2.3.2. Elastoplastic model of kernel

To better describe the mechanical properties of the kernel during the compression test, an elastoplastic model was applied based on Hertz contact mechanics (Yuan et al., 2022).

According to the Hertz contact theory:

$$F=\frac{4}{3}{E}^{\ast }{R}_e^{1/2}{D}^{3/2}$$ (1)

Where $F$ is the contact force, N; $D$ is the deformation, mm; $E^{\ast }$ is the comprehensive elastic modulus, MPa; and $R_e$ is the equivalent radius, they are defined as:

$$\frac{1}{E^{\ast }}=\frac{1-{\mu }_1^2}{E_1}+\frac{1-{\mu }_2^2}{E_2}$$ (2)

$$\{\begin{array}{c}\frac{1}{R^{\prime }}=\frac{1}{R_U^{\prime }}+\frac{1}{R_L^{\prime }}\\ \frac{1}{R^{″}}=\frac{1}{R_U^{″}}+\frac{1}{R_L^{″}}\\ R_e={\left(R^{\prime }{R}^{″}\right)}^{1/2}\end{array}$$ (3)

Where $E_1,E_2,{\mu }_1,{\mu }_2$ are the elastic modulus and Poisson ratio of the two contact objects respectively; $R_U,R_U^{\prime }$ is the curvature radii of the upper plate contact surface, minimum and maximum respectively, mm; $R_L,R_L^{\prime }$ is the curvature radii of the lower plate contact surface, minimum and maximum respectively, mm. The kernel apparent modulus of elasticity $E_1$ was decided by:

$$E_1=\frac{0.338F\left(1-{\mu }^2\right)}{D^{3/2}}{\left[K_U{\left(\frac{1}{R_U}+\frac{1}{R_U^{\prime }}\right)}^{1/3}+K_L{\left(\frac{1}{R_L}+\frac{1}{R_L^{\prime }}\right)}^{1/3}\right]}^{3/2}$$ (4)

Where D is the displacement, mm; $\mu$ is Poisson's ratio and set as 0.40 (Davison et al., 1975); The constants $K_U,K_L$ are determined according to the table provided in (ASAE, 2022). The value of $\mathit{cos}\theta$ was calculated by:

$$\mathit{cos}\theta =\frac{\frac{1}{R_U}-\frac{1}{R_U^{\prime }}}{\frac{1}{R_U}+\frac{1}{R_U^{\prime }}+\frac{1}{R_L}+\frac{1}{R_L^{\prime }}}$$ (5)

With the increased external force applied on the kernel, its structure began to deform plasticly due to the initiation, expansion, and fusion of micro-defects. To describe this behavior, a linear strengthening (softening) elastoplastic model was applied, as used for both rock and rapeseed compression processes (He et al., 2008; Yuan et al., 2022). The relationship between the contact force $F$ and the displacement $D$ was:

$$F=\{\begin{array}{c}\frac{4}{3}{E}^{\ast }{R}_e^{\frac{1}{2}}{D}^{\frac{3}{2}}\\ F_s+k\left(D^{\frac{3}{2}}-D_s^{\frac{3}{2}}\right)\end{array}\begin{array}{l}D\le D_s\\ D\ge D_s\end{array}$$ (6)

Where $F_s$ is the critical elastoplastic force; $D_s$ is the critical elastoplastic deformation; and $k$ is the strengthening (softening) coefficient. The values of $k$ and $E^{\ast }$, which dominant the mechanical properties of rapeseed, vary depending on the cultivar, moisture content, and pretreatment method.

Stress $\sigma$ was calculated as:

$$\sigma =\frac{F}{A}$$ (7)

Where $A$ is the contact area, assumed to be a sphere, and is given by:

$$A=\pi \left(2D-D^2\right)$$

Where $F$ is the contact force, $D$ is the displacement.

Since the diameter of the kernel was 2 mm, the strain $\epsilon$ was defined as:

$$\epsilon =\frac{D}{2}$$ (8)

Since the kernel breaks earlier than the whole seed, the extended kernel property regression was used as a reference when simulating the compression test for the whole seed.

2.3. 2D compression model

A 2D finite element model was developed in Abaqus/Standard 2020 (Dassault Systèmes Simulia Corp., Johnston, RI, USA), to characterize the mechanical properties of a canola seed, comprising both hull and kernel, under a compression test. Seeds were positioned between a movable top plate and a fixed bottom plate (Fig. 2 A). The hull and kernel were considered as distinct components, each assigned with specific material properties. To ensure a higher-quality mesh, both components were partitioned into several regions based on their geometry, facilitating the generation of a structured mesh using 4-node quadrilateral (CAX4R) elements. The plates were modeled as rigid bodies, assuming no deformation. The geometric parameters are provided in Table 1.

Table 1.

Geometric parameters of the 2D compression model.

Details	Parameters
Kernel radius (mm)	1
Hull inner radius (mm)	1
Hull outer radius (mm)	1.035

For validating the finite element model, the kernel mechanical properties, which derived from experimental compression tests, were firstly input into the kernel-only model. The resulting force and displacement (F-D) curve from the 2D model was compared with experimental compression test, an agreement between the two confirmed the validity of the kernel model. Subsequently, the validated kernel properties were incorporated into a comprehensive 2D finite element model that included both the hull and the kernel. The material properties of the hull were obtained from (Davison et al., 1975). The simulated F-D curve of the whole canola seed under compression was compared with experimental results for further validation of hull properties.

2.4. Static friction coefficients

For better dehulling performance, rubber was adopted and coated on the roller because it has the highest friction coefficient with rapeseed compared with the other materials (Izli et al., 2009). To investigate the impact of the pretreatment process on the interaction between seed and rubber surface, one group of canola seeds was prepared to 5.8 %, 16 % and 25 % (w.b.) by adding water, defined as rewet (RW). Another group was prepared using the FBD to 5.1 %, 16 %, and 26 % (w.b.), defined as wet and heat (WH). For each moisture content in both pretreatment methods, 30 seeds with a diameter of 2 mm were selected and tested on the lab-designed rubber-coated steel board. The coated rubber had a thickness of 0.3 mm (Plasti Dip ®Aerosol Spray). Statistical analyses were performed using OriginPro 2022 (OriginLab Corporation, Northampton, MA, USA). Differences were considered statistically significant at a 95 % confidence level ($p<0.05$).

2.5. 3D dehulling model

To simulate the mechanical process of the practical dehulling experiment, a 3D dehulling model was established with Abaqus/Explicit 2020 (Dassault Systèmes Simulia Corp., Johnston, RI, USA), with geometric nonlinearity (NLGEOM) enabled. The model consisted of a pair of rollers with a canola seed dropped between them (Fig. 2 B). The interaction of different mechanical loading on the seed were compared. The canola seed model considered the hull and kernel as independent component with their own material properties. Both are assumed to be isotropic materials and in a perfect spherical shape. The hull was simplified as an annulus covering the kernel to illustrate the impact of the mechanical force on the hull and its subsequent movement tendency during the process. The rollers were set as rigid bodies in the model as they were made up of steel. To improve the mesh quality and ensure accurate stress calculations, both the hull and kernel components were first partitioned into structural regions and then structurally meshed using 8-node hexahedral (C3D8R) elements. Table 2 presents the main geometric and mechanical parameters of the 3D model (Martinez-Soberanes et al., 2022).

Table 2.

Main geometric and mechanical parameters of the dehulling model.

Details	Parameter
Roller inner radius (mm)	6
Roller outer radius (mm)	25.1
Roller thickness (mm)	5
Hull inner radius (mm)	1
Hull outer radius (mm)	1.035
Hull annulus covering angles (degree)	20
Kernel radius (mm)	1
Hull Poisson's ratio	0.25
Kernel Poisson's ratio	0.25
Friction coefficient	0.27
Hull Young's modulus (MPa)	1027
Kernel Young's modulus (MPa)	499
Gap between the rollers (mm)	2
Left roller rotating speed (rpm)	669
Right roller rotating speed (rpm)	955

2.6. Dehulling experiment

The effects of the two dehulling modes—compression and shear—on seed deformation and fracture behavior were investigated. Dehulling experiment was recorded by high-speed camera, and synchrotron-based X-ray imaging was employed to visualize the complex internal structure of canola seeds following dehulling. The experiments were performed using a laboratory-designed roller dehulling machine equipped with a pair of rollers operating under compression and shear modes, respectively. The gap between the roller was set as 1.60 mm, and the roller speed was 112 rpm and 96 rpm. These parameters were optimized to achieve the most effective dehulling performance, as determined in (Martinez-Soberanes, 2023).

2.6.1. High-speed camera recording

Raw seeds at 5.4 % (w.b.) and FBD-pretreated seed at 16 % (w.b.), both with a 2 mm diameter, were dehulled by roller machine under compression and shear mode respectively. A CHRONOS 1.4 high-speed camera equipped with Computar 12.5–75 mm f/1.2 zoom lens (Kron Technologies Inc. 113–8337 Eastlake Drive Burnaby, BC, V5A 4W2 Canada) was used to record the dehulling experiment. The camera was configured with a resolution of 912 $\times$ 912 pixels, a frame rate of 1670 fps, and an exposure time of 595 us.

2.6.2. Micro-tomography technology

The internal morphology of the dehulled canola seeds subjected to compression and shear modes was examined at the BMIT-BM beamline in the Canadian Light Source (Saskatoon, Canada). Seeds were securely mounted on a holder and scanned by the synchrotron-base X-ray. The acquired image datasets were reconstructed and processed into 3D volumetric models using Avizo 3D 2022.2 software (Thermo Fisher Scientific, Waltham, MA).

3. Results

3.1. Compression behavior on different moisture contents

To establish how fluidized-bed drying (FBD) pretreatment influences the mechanical response required for controlled fracture during dehulling, compression tests were conducted with the raw canola seed and FBD-pretreated seed on different moisture contents (Fig. 3). Raw seeds exhibited higher rupture forces and limited deformation due to the high rigidity. FBD pretreatment modified the seed's mechanical properties, reducing rigidity and enabling controlled deformation prior to rupture.

Fig. 3.

Force-displacement curves of whole seed at different moisture contents under compression test.

The force–displacement curve showed two distinct deformation phases: an initial linear region corresponding to the elastic deformation of the kernel–hull assembly, followed by a nonlinear region reflecting progressive kernel deformation, internal hull expansion, and accumulated tensile stress that ultimately resulted in hull rupture (Haman et al., 1994). As moisture content increased, both rupture force and displacement at failure increased, indicating enhanced kernel plasticity and hull flexibility. These changes might allow the kernel to absorb and redistribute more deformation energy without fracturing, while the softened and stretchable hull experienced localized stress concentration along the interface. The resulting mechanical mismatch between a ductile, energy-absorbing kernel and a flexible yet weakened hull redirected stress. Consequently, cracks preferentially initiated and propagated along this interface, forming a controlled fracture pathway that enabled selective hull rupture while preserving kernel integrity.

3.2. Compression characteristics for kernel

Understanding the kernel deformation behavior is essential for preventing internal failure during dehulling. Kernel compression exhibited two characteristic stages separated by the elastoplastic deformation $D_s$: an initial elastic region, where deformation was recoverable upon unloading, followed by a plastic region in which increasing force led to the accumulation of internal micro-defects, progressive softening, and ultimately kernel crushing (Fig. 4 A).

Fig. 4.

Comparison of force–displacement (F–D) curves from experimental tests and finite element (FE) simulations. (A) Compression behavior of the canola kernel: experimental F–D curve overlaid with the simulated response and fitted elastoplastic model, all exhibiting two distinct deformation regions—an initial elastic region followed by a plastic region. (B) Compression behavior of the whole canola seed (kernel + hull): experimental F–D curve compared with the 2D FE simulation, validating the mechanical properties of both components.

The experimentally observed deformation pattern was captured by the elastoplastic model, which reproduced the kernel deformation and provided a theoretical basis for determining mechanical design parameters essential for controlled dehulling of canola. It was observed that in the first stage of compression ($D\le D_s$), the force–displacement (F-D) curve derived from Eq. (6) was approximately linear and in good agreement with the experimental results. In the second stage, the critical elastoplastic force ($F_s$) was incorporated into Eq. (6) to estimate the plastic softening coefficient $k$, calculated as the average of 10 randomly selected points from the region where deformation exceeded $D_s$. At 16 % (w.b.), the elastic modulus $E_1$ of the kernel was determined to be 30.5 MPa, while the elastic modulus $E_2$ of the flat stainless compression head was 7 $\times {10}^4$ MPa with Poisson ratio ${\mu }_2$ = 0.25 (ASAE, 2022). The critical elastoplastic deformation ($D_s$) was 0.10 mm, and the plastic softening coefficient $k$ was 9.00.

To validate the 2D finite element (FE) model of kernel, the stress and strain relationship was derived from the F-D curve of the compression test and determined using Eqs. (7) and (8). These parameters were then incorporated into the FE model to simulate the kernel deformation behavior. The F-D curves obtained from both the simulation and the experimental tests were plotted in Fig. 4 A. Moreover, the simulation model enabled estimation of the kernel ultimate stress, which could not be directly measured by mechanical testing. The ultimate stress values were determined as 3.9 MPa, 2.8 MPa, and 0.44 MPa for kernels at 8.8 %, 16 %, and 26 % (w.b.), respectively (Table 3). These findings confirm the reliability and applicability of the elastoplastic and FE models in characterizing the mechanical behavior of canola kernels under compression.

Table 3.

Mechanical properties of kernel and hull under different moisture contents.

	Kernel					Hull
Moisture Content (w.b.)	$E_1$ (MPa)	$E^{\ast }$ (MPa)	Elastoplastic deformation $D_s$ (mm)	$k$	Ultimate stress ($N/{mm}^2$)	Elasticity Modulus (MPa)	Ultimate stress ($N/{mm}^2$)
8.8 %	104	124	0.055	35.2	3.9	8021	56
16 %	30.5	74.2	0.10	9.00	2.8	1500	11
26 %	3.67	4.37	0.20	1.76	0.44	1250	24

3.3. Compression characteristics for hull

With the establishment of kernel mechanical properties, the mechanical properties of hull were subsequently investigated. A 2D finite element model consisting of both the hull and kernel was developed. The stress and strain relationship of kernel were obtained from the kernel compression tests, whereas the corresponding relationship for the hull were referenced from Davison et al. (1975) and iteratively adjusted in the simulation until a reasonable agreement was achieved between the simulated and experimental compression test (Fig. 4 B). The resulting elastic modulus and ultimate stress values for the hull were presented in Table 3.

Previous studies have reported a decreasing trend in the elastic modulus of the hull with increasing moisture content, which is consistent with our findings. However, due to differences in pretreatment methods, the elastic modulus of the hull obtained in this study was significantly higher than the values reported by other researchers, which ranged from 120 to 4027 MPa (Davison et al., 1975; Martinez-Soberanes et al., 2022; Dobrzanski and Stepniewski, 1991).

The dehulling experiments were then conducted to examine how these mechanical properties influenced seed behavior during dehulling. Canola seeds with moisture contents of 8.8 %, 16 %, and 26 % (w.b.) were dehulled using a roller machine operating in shear mode. For each moisture level, 10 seeds with an average diameter of 2 mm were processed in triplicate. The process yielded an average of 1, 5, and 0 intact hull-free kernels, respectively. At 8.8 % (w.b.), most kernels fractured due to their brittleness and limited ability to absorb stress. Conversely, at 26 % (w.b.), the hull and kernel remained attached as the seeds twisted, suggesting excessive moisture compromised effective hull detachment. The optimal performance occurred at 16 % (w.b.), where the hull exhibited the lowest ultimate stress of 11 MPa and the kernel reached an elasticity modulus of 30.5 MPa. This mechanical balance allowed the hull to bear most of stress and rupture while the softer, more compliant kernel absorbed and redistributed stress, facilitating separation without significant damage to the kernel.

3.4. Static friction coefficient

The static friction coefficient shown in Fig. 5 increased with moisture content for both pretreatment methods, which is consistent with previous reports for rapeseed (Çalışır et al., 2005; Izli et al., 2009; Xu et al., 2019). As moisture content increases, water absorbed by the porous hull structure promotes adhesion at the seed–surface interface and enhances viscoelastic deformation of the hull, both of which increase resistance to sliding (Stanley et al., 1976).

Fig. 5.

Static friction coefficient of canola with rubber surface from two pretreatment methods: wet and heat (WH) by fluidized bed dryer, rewet (RW) by adding water.

At same moisture content, FBD-pretreated (WH) seeds consistently exhibited lower friction coefficients than rewet (RW) seeds. This difference is attributed to pretreatment-induced changes in moisture distribution and hull microstructure. Rewetting primarily increases hull and near-hull moisture, leading to higher local adhesion and surface tackiness. In contrast, fluidized-bed drying redistributes moisture more uniformly within the seed and involves thermal exposure that can modify hull surface structure, thereby reducing effective surface adhesion. Heat-assisted treatments have been reported to induce microstructural rearrangement of plant hull materials, resulting in increased surface disorder and reduced surface continuity compared with untreated samples (Agu et al., 2017). These effects reduce effective contact adhesion and friction during roller interaction, thereby influencing the overall separation efficiency and performance.

3.5. Finite element simulation of dehulling mechanics

3.5.1. Compression mode

The finite element (FE) simulation elucidated the stress evolution and fracture mechanism governing the separation between the hull and kernel under compression mode (Fig. 6). When the seed first contacted the roller (Figs. 6 A–1), compressive stress was concentrated on the outer hull surface at the contact region, while tensile stress developed along the inner layer due to radial stretching. As the seed went deeper into the roller gap (Figs. 6 A–2), both the magnitude and area of compressive stress increased, progressively penetrating toward the inner hull layer. Simultaneously, tensile stress intensified around the contact zone due to two main effects: 1) the Poisson-induced lateral expansion of compressed regions, constrained by the rollers, generated localized tensile stress; and 2) frictional forces between the roller and hull that generated tangential tension. At the narrowest roller gap (Figs. 6 A–3), compressive stress intensified and transmitted inward through the hull, while the kernel exerted counter-pressure that redistributed stress across the interface. This interaction expanded the tensile zones along the hull edges, producing outward bulging and strain accumulation. Upon release (Figs. 6 A–4), tensile stress dominated the outer hull, confirming that fracture initiation was tensile-driven. The maximum compressive stress consistently appeared at the outer hull due to direct roller contact, creating localized high-pressure points that triggered micro-crack formation.

Fig. 6.

Contour diagram illustrating the mechanical response of canola seed under compression mode during dehulling. (A) Maximum principal stress of hull (MPa), positive value indicates tension stress, negative value represents compression stress. (B) Plastic deformation (PEEQ) of hull, showing deformation progression. (C) Maximum principal stress of kernel located at narrow gap between the roller, positive value indicates tension stress, negative value represents compression stress.

Plastic deformation analysis (Fig. 6 B) showed yielding originated in the inner hull layer and expanded outward, indicating fracture propagation from inside to outside under tensile stress. This inward-to-outward propagation was irregular, reflecting an uncontrolled fracture pathway rather than the guided interfacial separation desired for selective dehulling.

The kernel, protected by the hull, experienced secondary stress transmission rather than direct fracture (Fig. 6 C). Compressive stress localized at the outer kernel surface, surrounded by tensile zones (Figs. 6 C–1). In the cross-section along the loading direction (Figs. 6 C–2), the compressed region narrowed as stress intensified, leading to the gradual emergence and coalescence of tensile zones around it. Most of the kernel structures underwent tensile stress, resulting in localized expansion and bulging. This inward stress penetration indicates partial energy transfer from the hull into the kernel, creating a risk of internal damage—an outcome undesirable for achieving controlled dehulling, where the applied stress should be confined to the hull to ensure selective rupture and kernel preservation.

3.5.2. Shear mode

Under shear loading, the FE simulation revealed a similar stress evolution pattern but with greater directional control over fracture propagation (Fig. 7). When the seed first contacted the roller (Figs. 7 A–1), normal compression was concentrated on the outer hull surface, while tensile stress developed on the inner layer. As interaction continued (Figs. 7 A–2), the compressed area and stress magnitude increased, forming a tensile band around the contact region. The opposing roller rotation induced tangential shear forces that rotated the seed, stretching the hull laterally and initiating micro-sliding at the hull–kernel interface. As the seed entered the roller gap (Figs. 7 A–3), the upward motion of the left roller and downward motion of the right roller imposed vertical shear, twisting the seed between the rollers and amplifying tensile stress above and below the contact zone. This vertical shear transformed the earlier tangential tension into a fracture-driving force, directing crack propagation along the hull membrane rather than through the kernel. Upon release (Figs. 7 A–4), tensile stress was widely distributed along the hull circumference, promoting significant deformation and loosening of the hull.

Fig. 7.

Contour diagram illustrating the mechanical response of canola seed under shear mode during dehulling. (A) Maximum principal stress of hull (MPa), positive value indicates tension stress, negative value represents compression stress. (B) Plastic deformation (PEEQ) of hull, showing deformation progression. (C) Maximum principal stress of kernel located at narrow gap between the roller, positive value indicates tension stress, negative value represents compression stress.

Plastic deformation analysis (Fig. 7 B) showed tensile initiation at the inner hull, which expanded outward under the influence of rotational shear. This cumulative deformation weakened hull stiffness and created a directional fracture pathway, where the combined tangential and vertical shear stresses peeled the hull outward and forward in a progressive manner, enabling controlled detachment rather than random breakage.

The kernel exhibited asymmetric stress patterns due to the opposing roller motions. On the left side, in addition to the compression stress at the contact areas on both ends, a tensile stress was observed above the contact region. This was attributed to the upward friction generated between the hull and the kernel. As shown in Figs. 7 C–1, with deeper seeing into the kernel, the compressed area diminished while the compression stress intensified, reaching its maximum before being gradually overtaken by the increasing tensile stress. Conversely, on the right side (Figs. 7 C–2), the downward rotation of the seed caused friction between the hull and kernel, inducing tensile stress below the contact area. As seeing deepened, both the compression and tensile stresses decreased, with the compression stress eventually dissipating entirely. The overall stress magnitude within the kernel was lower than in compression mode, indicating reduced susceptibility to internal damage. These results demonstrate that shear loading enhances interfacial sliding and guides fracture along the hull surface, achieving efficient hull removal with minimal kernel injury.

3.5.3. Comparison of compression and shear mode

The FE simulations highlighted fundamental differences in stress transmission, fracture evolution, and kernel response between the two dehulling modes (Fig. 8). In compression mode (Fig. 8 A), stress rapidly intensified near the contact area, causing hull cracking but limited interfacial separation. The fracture path remained confined and discontinuous, leaving the hull tightly adhered to the kernel after unloading.

Fig. 8.

FE model for canola seed in the mechanical dehulling process of (A) Compression mode and (B) Shear mode.

In contrast, under the shear mode (Fig. 8 B), stress was distributed over a broader region and persisted longer as the seed rotated between rollers. The processing duration was approximately four times longer, allowing progressive interfacial weakening and hull peeling. Tensile and shear stresses concentrated along the hull edges, guiding fracture propagation around the kernel rather than through it. Consequently, the hull ruptured and peeled away from the kernel requiring less force. These observations align with the findings of Martinez-Soberanes et al. (2022), who reported that the ultimate strength of canola seeds at constant moisture content is lower under shear than under compression.

Kernel integrity also differed notably between the two modes. In shear mode (Figs. 7 C–2), maximum tensile stress was limited to the outer kernel surface, implying that potential oil loss would be limited to superficial layers while the internal structure remained intact. Conversely, in compression mode (Figs. 6 C–2) maximum tensile stress penetrated deeper into the kernel, increasing the likelihood of structural fragmentation.

Overall, the FE analysis demonstrates that compression mode induces stress concentration and uncontrolled fracture, while shear mode distributes stress directionally, enabling selective hull removal through a guided fracture pathway that preserves kernel integrity.

3.6. Dehulling experiment and microtomography validation

To validate the mechanistic hypothesis of selective fracture propagation, dehulling tests were conducted under compression and shear modes, and the resulting seed morphology was analyzed by high-speed imaging and synchrotron X-ray microtomography.

Control seed at 5.4 % (w.b.) fractured catastrophically under both modes (Fig. 9 A). In compression mode, the seed was clamped by the rollers and moved downward by static friction. As the rollers pressed on the seed surface, the applied normal force generated intense internal tensile stresses that quickly propagated through the hull-kernel structure. However, the rigid structure and the tight connection between the hull and kernel allowed little room for deformation. This intensified force transmission, causing the seeds to break into multiple fragments rather than controlled hull cracking. Similarly, in the shear mode, the accumulated tension stress derived from the compression ruptured the seed, but the shear force from different directions tore the seed into large fragments, again indicating an uncontrolled fracture pattern.

Fig. 9.

Experimental and imaging results comparing compression and shear dehulling modes for control and pretreated canola seeds. (A) Visual comparison of seed integrity and hull removal under compression and shear modes for raw (5.4 % w.b.) and pretreated (16 % w.b.) seeds. (B) Transverse slice and 3D reconstructions (front and back views) of a pretreated seed after dehulling in compression mode, showing partial hull removal and kernel deformation. (C) Transverse slice and 3D reconstructions (front and back views) of a pretreated seed after dehulling in shear mode, showing complete hull removal and preserved kernel structure.

In contrast, FBD-pretreated seed at 16 % (w.b.) exhibited a distinctly controlled deformation response. The softened, more elastic kernel redistributed stresses through shape adjustment, while the interfacial air gap created during pretreatment provided space for local deformation and guided crack propagation along the hull–kernel interface. Under compression mode, the hull cracked but remained attached to the kernel, indicating incomplete separation due to localized normal stress and limited lateral displacement. Notably, under shear mode, the combination of tangential and normal forces produced directional stress gradients that propagated cracks along the hull surface, leading to complete hull detachment and an intact, spherical kernel. The process occurred in two sequential steps: (1) compression increased internal pressure and initiated micro-cracks along the weakened hull; (2) the counter-directional shear forces rotated the seed, extending these cracks circumferentially until the hull peeled off completely. This two-stage fracture progression demonstrates the designed selective pathway—controlled hull rupture with kernel preservation.

Synchrotron X-ray microtomography further revealed distinct internal structural outcomes between the two modes (Fig. 9B and C). In compression mode, stress localization beneath the rollers caused kernel indentation and partial hull separation, confirming the destructive stress concentration predicted by FEM. In contrast, shear loading distributed stresses multi-directionally around the seed, promoting gradual hull peeling and preventing kernel cracking. The resulting 3D reconstructions clearly showed a continuous fracture plane along the hull–kernel interface, confirming the engineered fracture pathway predicted by simulation.

Together, these results validate that controlled pretreatment and shear-dominant mechanical loading establish a selective fracture mechanism that enables clean hull removal while maintaining kernel integrity.

4. Conclusions

This study established a mechanistic framework for achieving fracture-controlled canola dehulling by integrating fluidized-bed drying (FBD) pretreatment, finite element (FE) modeling, and experimental validation. FBD pretreatment effectively tuned the mechanical contrast between the hull and kernel—reducing hull rigidity and increasing kernel elasticity—thereby creating a stress differential that promoted preferential fracture along the hull–kernel interface. At the optimal moisture content of 16 % (w.b.), the hull reached its lowest ultimate stress (11 MPa) while the kernel retained sufficient elasticity (30.5 MPa), enabling the hull to rupture while the kernel absorbed deformation energy without damage.

The 3D FE model revealed that compression mode produced localized stress concentration and uncontrolled fracture, often transmitting stress into the kernel. In contrast, shear mode generated a combination of tangential and vertical shear forces that distributed stresses circumferentially, guiding crack propagation along the hull membrane. This directional stress field established a controlled fracture pathway, promoting gradual hull peeling rather than random breakage. Synchrotron microtomography validated these predictions, showing continuous interfacial separation and intact kernels under shear loading.

Together, the results demonstrate that coupling FBD pretreatment with shear-dominant mechanical loading enables selective hull removal through guided interfacial fracture, minimizing kernel loss and providing a scalable strategy for precision dehulling of small oilseeds. This mechanistic understanding of stress transfer, deformation, and fracture propagation establishes a scientific foundation for designing precision-controlled, energy-efficient dehulling systems. Beyond canola, these insights extend to a wide range of oilseed and biomaterial processing applications, supporting the development of next-generation mechanical separation technologies that preserve structural integrity, enhance product quality, and improve resource efficiency across the food and bioprocessing industries.

Author contributions

Runrong Yin: Conceptualization; Methodology; Data curation; Formal analysis; Investigation; Writing– original draft.

Edgar E. Martinez-Soberanes: Conceptualization; Methodology.

Wenjun Zhang: Conceptualization; Resources; Supervision; Writing – review & editing.

Martin J. T. Reaney: Conceptualization; Funding acquisition; Resources; Supervision; Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Handling Editor: Professor Alejandro G.Marangoni