Improving nutrient utilization in ruminants through extruded pelleted feeds

1. Introduction

High-producing dairy cows have high nutrient requirements, commonly met by feeding concentrated feeds rich in nutrients that can be digested and absorbed as volatile fatty acids in the rumen or as glucose and amino acids in the small intestine. Starch and protein are essential for rumen microbial growth, and a synchronized release and digestion of these nutrients in the rumen can enhance microbial protein synthesis and thus amino acids absorbed in the small intestine. However, rumen fermentation is complex and a balance between rapid and more slowly digested nutrients is crucial for regulation of feed intake and thus nutrient uptake (Allen, 2000; 2023) without digestive disturbances like rumen acidosis that can be detrimental for nutrient utilization and gut health, and thus animal well-being and production (Plaizier et al., 2022). Moreover, in high-yielding cows, starch is more energetically efficient when absorbed as glucose in the small intestine rather than fermented in the rumen, and enhanced amino acid absorption from rumen-escaped protein may similarly benefit performance (Owens et al., 1986; Schwab and Broderick, 2017).

In this context, the balance between rumen digestion and escape of starch and protein is of paramount importance. Rumen digestion is a competition between rate of passage and rate of digestion. The rate of passage determines how long a nutrient remains available for digestion, whereas rate of digestion reflects how rapid degradation occurs (Mertens, 2005). Together, these factors determine the extent of rumen digestion, or conversely, the extent of nutrient escape from the rumen. Factors affecting the rate of rumen digestion have been intensively studied, and knowledge on ingredient differences and effects of processing on the rate of rumen starch and protein digestion has been used to design diets for ruminants for decades. In contrast, the role of ingredient composition and processing in modulating the rate of rumen passage has received far less attention. In particular, the optimal processing conditions for producing pellet properties that facilitate rumen escape remain unclear. Rumen escape takes place through the omasal orifice in the reticulum. The presence of feed particles in the reticulum is mainly dependent on their specific density (Lechner-Doll et al., 1991; Offer and Dixon, 2000). Using plastic particles, desBordes and Welch (1984), Ehle and Stern (1986), and Murphy et al. (1989) demonstrated that high-density (1.17–1.42 g/mL) particles have approximately five times higher probability of escaping the rumen than low-density (< 1 g/mL) particles. These findings have been corroborated more recently by Seyama et al. (2017) and Dufreneix et al. (2019). The latter reported mean retention times in the rumen ranging from 8.9 h at 1.1 g/mL to 39.6 h at 0.9 g/mL density, corresponding to passage rates of 11.2 %/h and 2.5 %/h, respectively. Assuming a constant degradation rate of 10 %/h and first-order kinetics, increasing the passage rate from 5 % to 10 %/h increases rumen escape by 50 %, from 333 to 500 g/kg DM. Thus, although it must be verified using real feeds, the study of Dufreneix et al. (2019) indicates a substantial potential for influencing rumen escape of nutrients provided a feed particle with an optimal density for rumen escape.

Optimizing the density of individual feed particles is challenging but agglomerating them into pellets may help to overcome this limitation. Unlike plastic particles, feed pellets are not inert and their density as well as structural integrity can change in the wet rumen environment, making pellet stability equally important for maintaining their desired density properties. Conventional pelleting methods typically yield high-density pellets with low water stability (Larsen and Raun, 2018), leading to increased degradation in the ventral rumen and reducing their potential for rumen escape. Although the inclusion of binders can enhance pellet integrity and stability (Saleh et al., 2021; Stojkov et al., 2025), achieving precise control over pellet density remains challenging. Extrusion offers an alternative approach: as a thermo-mechanical process, it cooks pre-conditioned feed material under high temperature, pressure, and shear, transforming it into a viscoelastic dough or melt that is then pushed through a die at the extruder outlet (Miladinovic and Zimonja, 2010). As the melt exits the die, steam flashes off, forming a porous and expanded pellet with distinct physical properties such as density, durability, hardness, sinking velocity (Sørensen, 2012; Khater et al., 2014), as well as improved water stability (Welker et al., 2018). Various feed formulations and extruder settings such as screw configuration, screw speed, temperature control in different sections of the extruder barrel, and die dimensions have been widely studied in aquaculture feed production to obtain pellets with desired characteristics (e.g., density and sinking or floating behavior) (Rolfe et al., 2000; Chevanan et al., 2007; Sørensen et al., 2010; Draganovic et al., 2011; Kraugerud et al., 2011; Fallahi et al., 2013; Welker et al., 2018). Based on these studies, low-density, floating pellets can be produced by increasing the extruder temperature through methods such as higher screw speeds (i.e., 300–500 rpm), direct steam injection, or heating the extruder barrel. To obtain high-density, sinking pellets an extruder screw speed around 200 rpm and controlling the extruder die temperature ≤ 100 ◦C by water cooling of the last section have been shown to be optimum (Wang et al., 2021). However, such investigations are limited in ruminant feeds.

In ruminants, water-stable feed pellets designed with high density to promote rumen escape or with lower density to allow gradual fermentation while floating in the dorsal rumen may enhance overall nutrient utilization. A major challenge, however, is producing pellets that consistently exhibit these desired properties. Larsen et al. (2019) have studied the effect of extruded pellet density and stability on rumen digestion dynamics, but they characterized their feed pellet properties using methods developed for aquaculture feeds. Since the rumen environment markedly differs from that of the sea, particularly in terms of temperature, fluid viscosity, osmolarity, microbial activity, and physical dynamics, there is a need to define and evaluate the physical properties of extruded feed pellets specifically for ruminants.

The objective of this study was to determine whether the extrusion process can be optimized to produce feed pellets with targeted density and stability characteristics that promote either enhanced rumen escape or gradual fermentation, as evaluated by their behavior in a simulated rumen environment using various in vitro setups. Barley, maize and solvent extracted soybean meal (SBM) were selected as feed materials as they show a range in contents of starch, fiber and protein, and represent three major ingredients used as animal feed. The range in composition, together with extrusion settings selected based on prior research and a preliminary trial with barley, was expected to yield pellets with varying density properties suitable for evaluating effects under rumen-like conditions. This study focuses on the technical details of extruder settings to identify factors required to produce pellets with the desired rumen density and fluid stability. Additionally, we explored the potential of using reliable proxies to assess these properties during production.

2. Materials and methods

2.1. Experimental design and description of treatments

Feed processing was conducted at the Center for Feed Technology (ForTek) ˆ at the Norwegian University of Life Sciences (NMBU), Ås, Norway. The experimental design was factorial with one experimental unit per combination (Fig. 1). Factors studied were feed material (described below), particle size through screen size (2 mm or 6 mm) in a hammer mill, extruder screw speed (210 rpm or 300 rpm), and level of cooling in the last section of the extruder barrel (Yes or No).

Barley, maize, and soybean meal (SBM; solvent extracted obtained from Denofa AS, Fredrikstad, Norway) were used either individually or in mixtures of barley+SBM (B+SBM; 50:50 as is basis) and maize+SBM (M+SBM; 50:50 as is basis) giving a total of five feed materials. A total of 40 extruder pelleted treatments were produced, eight for each of the five feed materials (Fig. 1). Due to the adverse quality of pellets obtained in pre-trial testing of 100 % SBM, 10 % maize was included in SBM treatments. Treatments were produced in two rounds, with single-ingredient treatments produced in the first round and mixture treatments produced in the second round.

2.1.1. Processing of extruded treatments

Barley, maize, and SBM were ground separately in a hammer mill (E-22115 TF, Münch-Wuppertal, Germany, 18,5 kW el.motor, 2870 rpm) using 2 mm and 6 mm screen sizes. These screen sizes were chosen to provide a clear contrast in particle size and represent typical ranges used in ruminant feed processing. Samples of ground material were collected for chemical and particle size analysis (Table 1). After grinding and batching (i.e., weighing a specific quantity of feed material for each treatment run), all individual ingredients and mixtures were passed through a twin shaft paddle mixer (Forberg AS, Larvik, Norway) for 120 s. Thereafter, the feed mash was pre-conditioned similarly for all feeds in a double conditioner (BCTC 10, Bühler, Uzwil, Switzerland) with a constant feeder rate of 100 kg/h. About 18 ± 1 kg/h moisture was added in the conditioner as liquid (58.7 ± 5 % of added H2O) and steam (41.3 ± 5 % of added H2O) to get a total moisture content of 270 ± 10 g/kg of feed mash and a temperature between 85 and 90 ℃. The preconditioned mash was extruded in a co-rotating twin-screw extruder (Bühler BCTG 62/20 D; 5 barrel’s sections) and pelleted using 6 mm die size (revolver die; six number of dies). Four extruder operating parameters were used. Those were low (210 rpm) and high (300 rpm) screw speeds combined with either cooling (to control exit temperature between 80 and 90 ℃) or without cooling in the last section (5th section) of the extruder barrel. These settings were selected to create a broad range of pellet densities based on prior research and pretrial extrusion with barley. The screw length was 1260 mm, and the screw configuration was 100R100–100R100–80R80–80R80-P120–60L20(90◦twist-off)- 80R80–80R80–80R80–80R80–80R80–60R60–60R60–60R60–60R60–60R60–20R60–40R60, where the first number represents pitch length, the second number length of the screw element, and R and L indicate forward and backward conveying direction. The P120 is a polygonal kneading element having forward conveying properties.

The four treatments within screen size and feed material were produced in continuous runs. Within the run, sufficient material was processed to ensure sampling of treatments during steady state conditions. Steady state conditions were defined as torque variations in the extruder being within ±3 Nm for a minimum of 60 s. To avoid contamination and carry-over between samples, feed material where then run for 25 min to achieve sufficient product material for treatment sampling. Before sampling, the extruded pellets were dried in a fluid bed continuous dryer (ForTek, ˆ NMBU) for 7–10 min at approximately 100 ℃ to achieve a final moisture content of approximately100 g/kg of pellets. Despite adjustments for differences in pellet expansion, minor variations in moisture content (2–3 %) were observed among treatments. After drying, about 10 kg sample within treatment condition was taken when the pellets were exiting the fluid bed dryer. The sample was then cooled in batch coolers (ForTek, ˆ NMBU) using ambient air for 20–30 min. After cooling, this sample was split into representative sub-samples for further analysis. While the feeder rate was kept constant for all treatments, we acknowledge that the actual extrudate production rate at the die exit may have varied slightly with screw speed and process conditions, although this was not specifically measured. Extruder data such as barrel temperature at section 3 (T3) and section 5 (T5), die pressure (DP), torque, and specific mechanical energy (SME) were recorded directly from the extruder control panel and are reported in Table 2. The SME values were initially recorded as power consumption (kW). To account for the baseline energy required to rotate the screws without material, zero-load SME was measured separately under the same screw configuration. Net SME was calculated by subtracting the zero-load SME from the production SME (both in kW), and the result was normalized to the feed rate to express net SME in Wh/kg, allowing comparison across processing conditions.

Fig. 1. Flow diagram describing feed treatments. Feed materials used were barley, maize, soybean meal (SBM) in the first round, whereas barley+SBM (B+SBM; 50:50) and maize+SBM (M+SBM; 50:50) were used in the second round. For each feed material, there were 8 treatments.

2.1.2. Chemical composition and particle size analysis of feed materials

Dry matter (DM) was determined by oven drying at 103 ◦C for 18 h. Starch was determined by enzymatic hydrolysis into glucose (McCleary et al., 1994). Nitrogen was determined according to Kjeldahl-N, AOAC Method 2001.11 (Thiex et al., 2002), and crude protein (CP) was calculated as N × 6.25. Neutral detergent fiber (aNDFom) was determined using ANKOM220 fiber analyzer (ANKOM, Technology, Fairport, NY, USA) with heat-stable amylase and correction for residual ash (Mertens, 2002). Acid detergent fiber (ADFom) was determined according to Method 973.18 (AOAC, 2000) with correction for residual ash. Crude fat was determined by Accelerated Solvent Extraction (ASE200, Dionex Corporation, Sunnyvale, CA, USA). The particle size of ground feed materials was determined as geometric mean diameter (GMD) with geometric standard deviation (GSD) based on the formula described by ASABE (2013). The results are presented in Table 1.

2.2. Analysis of physical properties of pellets

Pelleted feed samples collected were subjected to analyses of physical properties in the form of specific density (SD), specific density in rumen fluid (SDrf), sinking velocity (SV) and fluid stability index (FSI). Moreover, radial expansion (RE) and bulk density (BD) were investigated as substitutes for SD and FSI. All analyses were performed with some modifications based on procedures used in the fish feed industry and are described in detail below.

2.2.1. Specific density (SD)

Specific density, also known as geometric envelop density of pellets, was calculated using the following formula:

Where Wp is the weight of pellets and Vp is the volume of the same pellets. Pellet weight was determined using a laboratory scale (AG204 DeltaRange®, Mettler-Toledo GmbH, Greifensee, Switzerland). To accurately determine the volume of irregular-shaped pellets, the volumetric displacement method was used. The method was modified from Hwang and Yakawa (1980) by using a tapped density analyzer as described below.

In short, glass beads with 0.5 mm diameter were used as displacement medium in a 10 mL or 25 mL graduated glass cylinder depending upon size of the pellets. Five to ten pellets were randomly selected and weighed (Wp) together. A few pellets with visibly open pores were discarded. Then, the volume of glass beads without pellets (Vi) was measured in a graduated glass cylinder by tapping 100 times with AUTOTAP (AUTOTAP, Quantachrome Instruments, 1900 Corporate Drive, Boynton Beach, Florida, USA). After that, glass beads were taken out of the glass cylinder, and some were poured back making a thin layer in the bottom of the cylinder. A pellet was placed on top of the layer, and enough glass beads to cover the pellet were poured into the cylinder. This process was continued until all the weighed pellets were covered in glass beads. Finally, the remaining glass beads were poured on top, and the cylinder was tapped 100 times again to obtain the final volume (Vf) of glass beads with pellets (Fig. 2). The volume of pellets (Vp) was then calculated as:

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 2

Subsequently, the SD of pellets (in g/mL) was calculated using the previously described Eq. (1). For each sample, five measurements of SD were taken.

2.2.2. Specific density in rumen fluid (SDrf)

Specific density was also determined after immersion in rumen fluid to investigate the change in density of pellets. The rumen fluid was collected from two rumen-cannulated cows fed a standardized diet at the maintenance level as described by (Åkerlind et al., 2011). The rumen fluid was collected in prewarmed thermos flasks approximately four hours after the morning feeding. After mixing, the fluid was strained through a 200 µm mesh cloth (Nitex 03–200/47 SEFAR, Heiden, Switzerland).

Five to ten selected pellets were soaked in rumen fluid at 39 ℃ for 20 min. After soaking, pellets were gently placed on tissue paper to absorb excess water on the pellet surface. Thereafter, SD was determined as described above and denoted as “SDrf” i.e., SD of pellets in rumen fluid. Three measurements of SDrf were taken for each feed sample.

2.2.3. Sinking velocity (SV)

Sinking velocity of pellets was determined in rumen fluid in a 250 mL transparent glass cylinder (310 mm long and 35 mm inner diameter) with two fixed points marked 220 mm apart (Fig. 3). There was a 30 mm column of fluid above and below the marked points. The cylinder with the strained rumen fluid was placed in an incubator cabinet at 42 ℃ to ensure a rumen fluid temperature between 38 and 39 ℃. A lamp was placed behind the glass cylinder to illuminate the rumen fluid to ease pellet movement observation. Randomly selected pellets were then dropped one by one from a height of about 30 mm above the fluid surface, and SV was determined as mm/s by measuring the time elapsed to travel the distance of 220 mm, using a manually operated stopwatch. For each feed sample, 30 randomly selected pellets were tested.

A supply of rumen fluid was stored at 39 ℃, and rumen fluid was renewed in the glass cylinder after ten pellets measured. The density of rumen fluid was also measured at 39 ℃, and it remained constant at 0.988 ± 0.001 g/mL.

2.2.4. Fluid stability index (FSI)

Fluid stability was determined by modifying the water stability index (WSI) method of Baeverfjord et al. (2006), developed for testing fish feeds. The main modifications were rotational agitation, higher temperature, and ruminal fluid as a medium instead of tap water. Thus, the name fluid stability index (FSI) was given instead of WSI.

The FSI was determined using ball-shaped stainless-steel baskets (Anping Amma Filter Equipment Co., Ltd., Hengshui City, China) having an inner diameter of 58 ± 2 mm, filtered rumen fluid, and the DaisyII Incubator (ANKOM Technology, Fairport, NY, USA) (Fig. 4). The original size of the basket mesh was 0.7 mm. To enhance the removal of disintegrating particles, holes of 2 mm were made manually with an awl. About 5 g of pellets were weighed into a basket that was closed tightly using an attached clip. Rumen fluid was collected and processed as described for SV. Two liters of rumen fluid were poured into a daisy incubator glass jar, fitted with two small bulges to ensure baskets’ twirling during rotation. Three baskets carrying three different feed samples were placed in each glass jar and wholly immersed in rumen fluid. Then, glass jars were placed in the DaisyII Incubator at 39 ℃ and rotated at a speed of five rotations per min. Each basket twirled 10 times per min. After 90 min, baskets were removed from the glass jars. After removing excess fluid, the baskets were cleaned outside gently with tissue paper, weighed, and placed in an oven at 103 ℃ for 18 h. Pellet stability was calculated as dry matter retained after incubation in rumen fluid divided by dry matter before incubation. The FSI was calculated using the following equation:

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 3

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 4

where W1 and W2 are the dry weights of the pellets with the basket before and after incubation (90 min), respectively, and Wb is the dry weight of the empty basket. The FSI was determined in triplicate for each treatment and expressed on a dry matter basis as grams of retained pellet DM per kilogram of initial pellet DM (g/kg DM).

2.2.5. Radial expansion (RE)

Radial expansion was determined by measuring the diameter of randomly selected pellets (n = 30) at three different points by a digital vernier caliper. The average pellet diameter was used to calculate expansion (%) by the following formula:

where Dp is average pellet diameter, and De is the die size in the extruder.

2.2.6. Bulk density (BD)

Bulk density was determined as described by Sørensen (2012). In this method, pellets are poured into a one-liter tared steel cylinder without agitating. Excess pellets are removed by a scraper, gently pulling over the edge of the cylinder. The cylinder with pellets was then weighed, and BD (g/L) was measured three times for each feed sample.

2.3. Statistical analysis

Within treatments, each physical quality was measured several times, and the average values were used to run the statistical analysis. The MIXED procedure of SAS (2013) was used to evaluate the effect of factors on SD, SDrf, SV and FSI of pellets according to the following model:

where; Yijklm is the dependent variable; µ is the overall mean of the dependent variable; αi is the fixed effect feed material (i = 5); βj is the fixed effect of screen size in hammer mill (j = 2); γk is the fixed effect of screw speed (k = 2); δl is the fixed effect of level of cooling in the last section of extruder barrel (l = 2); ηm is the random effect of treatment number; eijklm is the random errors associated with observation ijklm. Two-way interactions were used for all main effects, whereas three-way interactions were initially included but due to non-significant effect removed from the model. The variance-covariance (VC) was used as a covariance structure for the random effect. The significance level of each factor was determined by Kenward-Roger denominator degrees of freedom approximation for the type III test of fixed effects resulting from the model where restricted maximum likelihood (REML) was used as estimation method. The least-square (LS) means ± standard error of the LS means (SEM) of main effects and two-way interactions involving feed material are presented. Multiple comparisons were tested using the PDIFF statement and considered significantly different at P < 0.05 and as a tendency at 0.05 < P ≤ 0.10.

In addition, the Pearson product-moment correlation procedure in SAS (2013) was used to check inter-relationships between independent and dependent variables and among dependent variables. The dependent variables were extruder process variables (T3, T5, DP, torque, and SME) and pellet physical properties (SD, SDrf, SV, FSI, RE and BD). Independent variables were processing conditions (factors) such as screen size in hammer mill, screw speed, and cooling in the last section of the extruder. Results are presented as correlation coefficients (r) and considered significant at P < 0.05 and as a trend at 0.05 < P ≤ 0.10.

3. Results

3.1. Process variables during extrusion

The applied extrusion conditions successfully generated clear contrasts in thermal conditions across treatments. Without cooling the extruder’s last section, T3 ranged from 107 to 130 ◦C and T5 from 102 to 133 ◦C whereas with cooling, T3 ranged from 103 to 110 ◦C and T5 from 79 to 86 ◦C (Table 2). For all feed materials, T3 was positively correlated with T5 (r = 0.893, P < 0.001, n = 40). Cooling applied in the last section of the extruder barrel was negatively correlated with T3 (r = -0.694, P < 0.001, n = 40) and T5 (r = -0.933, P < 0.001, n = 40). The correlation between screw speed and torque was negative for maize, SBM, and M+SBM (r = - 0.694, P < 0.001, n = 24). When cooling was not applied in last section of the extruder barrel, SME was positively correlated with screw speed (r = 0.459, P = 0.05, n = 16), and T3 and T5 (r = 0.874, P < 0.001, n = 16) for all feed materials except M+SBM. The SME was correlated with DP, negatively for maize (r = -0.670, P = 0.05, n = 8), but positively for all other feed materials (r = 0.770, P < 0.001, n = 32). Overall, SME was positively correlated with torque (r = 0.848, P < 0.001, n = 40).

3.2. Physical properties of pellets

3.2.1. Specific density (SD) and SD in rumen fluid (SDrf)

Barley and maize exhibited a wider range of SD values from 0.43 g/mL to 1.23 g/mL, whereas SD of SBM, B+SBM and M+SBM ranged from 0.89–1.18 g/mL (Table 3). Increasing screw speed from 210 to 300 rpm decreased SD in maize and barley but not in SBM, B+SBM and M+SBM (PFM×SS = 0.006; Table 4). By applying cooling in the last section of the extruder barrel, SD increased in maize, followed by barley, B+SBM and M+SBM (PFM×C < 0.001). The highest SD was obtained for maize and M+SBM. The SDrf values increased compared to their respective SD values and this increase in density was highest for pellets with initial low SD (Table 3; Fig. 5). The overall correlation between SD and SDrf was positive and high (r = 0.961, P < 0.001, n = 40). However, unlike SD, the SDrf was higher with 2 mm than 6 mm screen size and with 210 rpm than 300 rpm screw speed for maize, whereas SDrf of barley, B+SBM and M+SBM were not affected by screen size or screw speed (PFM×SH = 0.011; PFM×SS = 0.045; Table 4). For all feed materials except SBM, SD and SDrf were negatively correlated with SME (r = -0.489, P = 0.005, n = 32) and T5 (r = -0.816, P < 0.001, n = 32).

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 7

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 8

3.2.2. Sinking velocity (SV)

Sinking velocity of barley ranged from floating (0 mm/s) to slow sinking (20 – 41 mm/s) to fast sinking (74 – 112 mm/s) (Table 3). Maize was either floating (0 mm/s) or fast sinking (100 – 180 mm/s) (Table 3). For SBM, B+SBM and M+SBM, the SV of pellets ranged from 104 to 114 mm/s, 70–114 mm/s and 80–156 mm/s, respectively. Without cooling in the last section of the extruder barrel, maize gave only floating whereas barley yielded floating to slow sinking pellets. With application of cooling in the last section, SV increased for maize, barley, and B+SBM whereas SBM and M+SBM remained unaffected (PFM×C < 0.001; Table 4).

3.2.3. Fluid stability index (FSI)

The FSI of barley and maize ranged from 666 to 943 g/kg DM, whereas for SBM and mixtures, FSI ranged from 32 to 115 and 36–366 g/kg DM, respectively (Table 3). With increase in screen size in the hammer mill from 2 to 6 mm, FSI of B+SBM and M+SBM increased but remained significantly lower than maize and barley (PFM×SH = 0.021; Table 4). Application of cooling in last section of the extruder barrel increased FSI in maize but decreased it in M+SBM (PFM×C = 0.003). A positive correlation of SME and torque with FSI was seen (r = 0.649, P < 0.001, n = 32) for feed materials, except SBM. Similarly, a significant correlation (P < 0.05) between FSI and DP was observed. This correlation was positive for maize (r = 0.745, n = 8) and negative for barley (r = -0.838, n = 8), B+SBM (r = -0.717, n = 8) and M+SBM (r = -0.865, n = 8).

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 9

Improving nutrient utilization in ruminants through extruded pelleted feeds - Image 10

3.2.4. Radial expansion (RE) & Bulk density (BD) and their relationship with SD, SDrf, SV & FSI

The RE of barley and maize ranged from 13 % to 52 % and 17–117 %, respectively, with corresponding BD ranges of 429–627 g/L and 284–835 g/L (Table 5). In contrast, SBM, B+SBM and M+SBM exhibited the smallest variation in RE (4–20 %), with BD values ranging from 568 to 743 g/L. Pellets with BD below 469 g/L were observed to float. For all feed materials except SBM, both SD and BD were negatively correlated with RE (r = -0.803, P < 0.001, n = 32). The overall correlation between SD and BD was high and positive (r = 0.959, P < 0.001, n = 40); however, when only pellets with SD values between 1 and 1.1 g/mL were compared with their respective BD, the correlation was not significant (r = 0.289, P = 0.21, n = 16). Since SD was highly correlated with both SDrf and SV (r = 0.917, P < 0.001, n = 40), the relationships of SDrf and SV with RE and BD followed a similar trend to that of SD. When comparing barley and maize, we observed a significant negative correlation between FSI and RE for pellets produced without cooling the extruder’s last section (r = -0.828, P = 0.01, n = 8). Conversely, for B+SBM and M+SBM, the correlation between FSI and RE tended to be positive (r = 0.432, P = 0.094, n = 16). Similar correlations were observed when RE was replaced with BD.

4. Discussion

4.1. Behavior of extruded pellets in rumen like environment

In the rumen, feed particles are stratified based on their density, which influences their sedimentation or floatation behavior (Sutherland, 1988). Particles with low density (< 1 g/mL) float in the dorsal compartment of the rumen. Conversely, denser particles (> 1 g/mL) settle in the ventral compartment, with sinking velocities dependent on their specific densities. Studies using inert plastic particles have suggested a minimum SD of around 1.17 g/mL as necessary for enhanced rumen escape (desBordes and Welch, 1984; Seyama et al., 2017). In our study, only some maize-based treatments achieved this threshold. However, as stated earlier, feed pellets are not inert, and their apparent density (or functional specific gravity) can change upon entering the rumen. Studies investigating feed particle dynamics suggest that particle density initially increases due to hydration by saliva and rumen fluid (Hooper and Welch, 1985). However, subsequent gas bubble adhesion during microbial fermentation (Wattiaux et al., 1992) may increase buoyancy by creating upward force without altering the pellet’s intrinsic mass or volume. These opposing forces complicate predictions of in vivo behavior. Recognizing this, Dufreneix et al. (2019) proposed a more conservative target density range of 1.2–1.3 g/mL to ensure passage from the reticulorumen.

To approximate the hydration effect, we determined the SDrf of pellets after immersion in rumen fluid for 20 min. The observed increase in density after hydration was notably greater for pellets with initially low SD (Fig. 5). In contrast, high SD pellets exhibited minimal density changes, likely because reduced porosity limited water absorption. Although water absorption can theoretically cause pellet swelling and offset density increases (Piedecausa et al., 2009), previous studies (Chen et al., 1999; Vassallo et al., 2006) and our own observations showed that pellet weight increased without significant dimensional changes after immersion, supporting a net rise in density. A 20 min soaking period was chosen based on preliminary tests that showed pellets with an initial SD below 0.78 g/mL experienced the greatest density increases. Despite this increase, the final SDrf of these low-SD pellets remained below the density of rumen fluid (≈ 0.988 g/mL), suggesting they would likely remain afloat in vivo. In contrast, pellets with an initial SD above 0.78 g/mL and below 1 g/mL attained SDrf values exceeding the density of rumen fluid. Thus, to ensure that pellets exhibit floating behavior in the rumen, their initial SD should be maintained below 0.78 g/mL.

For pellets to achieve optimal densities for rumen escape (> 1.2 g/mL), an initial SD above 1.05 g/mL appears necessary, considering expected hydration dynamics. While pellets with initial SD near 1.0 g/mL might also achieve suitable densities over extended rumen residence, such outcomes are uncertain and likely variable due to complex dynamics of particle hydration and microbial activity in the rumen. Therefore, the density threshold of 1.05 g/mL was chosen conservatively to account for this variability, although higher initial SD is more desirable. It is important to note that, apart from internal pellet structure, feed material composition significantly influences water absorption rates and subsequent density changes (Ramanzin et al., 1994; Hemmingsen et al., 2008). In our study, barley-based pellets demonstrated notably higher water uptake compared to other feed materials, suggesting faster attainment of optimal density for rumen escape.

Sinking velocity, another critical measure influenced predominantly by pellets density (Chevanan et al., 2007; Sørensen, 2012), was determined to further confirm the sinking or floating behavior of pellets in the rumen. Although SV is routinely measured in aquafeed studies, with typical ranges between 55 and 155 mm/s for slow-sinking salmon feeds (Chen et al., 1999; Piedecausa et al., 2009), this metric has not previously been applied to ruminant feeds. Recognizing that fluid temperature and properties such as salinity, viscosity and surface tension significantly impact SV (Chen et al., 1999), we measured SV using rumen fluid at cow body temperature (39 ◦C). We categorized pellets as slow sinking (< 40 mm/s), fast sinking (70–120 mm/s), and very fast sinking (> 120 mm/s), based on observed patterns. Clear distinctions in pellet behavior were associated with SD ranges: floating (< 0.78 g/mL), slow sinking (0.85–0.90 g/mL), fast sinking (0.97–1.15 g/mL), and very fast sinking (> 1.15 g/mL). Nevertheless, caution must be exercised when interpreting these in vitro results, as the presence of fibers and complex rumen dynamics in vivo might alter pellet sinking behavior.

To effectively utilize the pellet density properties, maintaining adequate pellet stability in ruminal fluid over a certain time is essential. Additionally, rapidly disintegrating pellets may accelerate rumen fermentation, adversely affecting the rumen environment. Therefore, pellet stability is critical both for expressing density and for its nutritional implications. However, criteria for assessing pellet stability specifically for ruminant feeds are not well-established. Pellet stability is commonly assessed as WSI, originally developed for aquatic feeds (Baeverfjord et al. (2006). Using this method, Larsen and Raun (2018) found that the WSI of 24 steam-pelleted commercial compound feeds for dairy cows ranged from 21 to 198 g/kg DM after 120 min incubation. In contrast, Larsen et al. (2019) reported WSI values ranging from 467 to 980 g/kg DM for steam-pelleted treatments, except for 100 % maize, which had a WSI of 39 g/kg DM. For extruded pellets containing 100 % cereal grains, they observed an average WSI of 832 g/kg DM, whereas their extruded mixtures (50 % cereal grains + 50 % SBM) had an average WSI of 717 g/kg DM. These differences among studies likely reflect variations in pellet composition and processing conditions but could also stem from inability of the WSI method to fully capture stability characteristics specific to ruminant feeds. The WSI is influenced by several factors, including agitation intensity, temperature, and ionic concentration of the incubation medium (Obaldo et al., 2002). Our study improved upon traditional methods by employing ruminal fluid maintained at cow body temperature (39 ◦C) combined with higher agitation, conditions better approximating actual ruminal dynamics than conventional water stability test. According to our criterion (FSI > 800 g/kg DM after 90 min incubation), only pellets composed exclusively of cereal grains achieved high stability for both low-density floating and high-density sinking treatments. Despite methodological enhancements, our in vitro approach does not fully replicate the complexity of the rumen environment, particularly the stronger physical contractions and the presence of fibrous digesta, and disruption of pellets by chewing. Consequently, our measured FSI values may overestimate actual in vivo pellet stability. Additionally, since the mean retention time of feed particles or nutrients in the rumen typically spans several hours, these findings underscore the need for further validation through in vivo studies to more accurately assess rumen escape.

4.2. Factors affecting density and fluid stability of extruded pellets

Our study demonstrates that extruder processing can be used to produce feed pellets with the targeted density and rumen fluid stability. Among the variables evaluated, feed material composition and the level of cooling in the last section of the extruder barrel emerged as the most critical factors, with extruder screw speed being nearly as influential, while hammer mill screen size had a comparatively minor effect. The varying proportions of starch, protein, and fiber in feed ingredients impart unique rheological properties to the viscoelastic extrusion melt, affecting expansion and ultimately the structural and physical characteristics of the extruded pellets (Forte and Young, 2016; Kristiawan et al., 2016). Since cooling was identified as the most influential extrusion variable, we discuss the effects of other factors on pellet density separately for conditions with and without cooling to better illustrate their specific contributions.

Pellets produced without cooling the extruder’s last section exhibited lower SD values in barley and maize than SBM and mixture treatments. Starch undergoes several structural changes during extrusion, including gelatinization, melting, and fragmentation (Lai and Kokini, 1991). Gelatinized starch forms a matrix that traps water vapors and expands when the pressure drops at the die exit (Moraru and Kokini, 2003), thereby reducing density. High starch content in maize, combined with increased shear and elevated temperature (reflected in higher SME and T3 values) which favor starch gelatinization (Diosady et al., 1985), likely resulted in maize producing the lowest SD and floating pellets, regardless of hammer mill screen size or extruder screw speed. In barley, higher fiber content inhibits starch gelatinization by reducing water binding capacity and thereby limiting expansion (Robin et al., 2012). Consequently, barley produced pellets with a range of SD values—primarily influenced by screw speed—which resulted in behavior ranging from floating to slow sinking in rumen fluid. Nonetheless, despite differing pellet behaviors, both barley and maize showed reduced density with increasing screw speed. This aligns with previous studies linking the effect to elevated shear and temperature (Baik et al., 2004; Fallahi et al., 2013; Kirjoranta et al., 2016). While smaller particle size is known to promote melt homogeneity and friction, enhancing expansion and thus reducing density (Arˆeas, 1992; Desrumaux et al., 1998), no clear effect of hammer mill screen size on SD was found in our study. Thus, to produce floating pellets (SD < 0.78 g/mL), an optimal screw speed of 210 rpm—without cooling the extruder’s last section—was sufficient for maize, whereas barley required a screw speed of 300 rpm to achieve comparable flotation behavior under the tested conditions. Higher screw speeds may benefit barley further, but this remains to be verified.

In contrast to cereal grains, SBM and its mixtures consistently produced higher SD, sinking pellets regardless of screw speed. Protein transformations during extrusion are somewhat analogous to starch gelatinization (Forte and Young, 2016) but typically require higher temperature and shear forces (Guy, 2001; Kraugerud et al., 2011) than those reached under our conditions. Despite high starch content, mixtures behaved similarly to SBM. Starch-protein interactions during extrusion may form insoluble complexes that reduce the water retention and limit starch gelatinization (Allen et al., 2007). Moreover, reduced water retention may also lower shear by lubricating the screws and barrel wall, as reflected by lower T3 and SME values in the mixtures, particularly in M+SBM. As a result, SBM-based pellets exhibited limited expansion and transformation. Given that our conclusions regarding these mechanisms are based on indirect evidence, our interpretation remains tentative.

Cooling the extruder’s last section below 90 ◦C significantly increased SD in all feed material treatments, except SBM. It has been reported that decreased temperature near the die exit increases melt viscosity and reduces pellet expansion (Akdogan, 1996; Suknark et al., 1999; Welker et al., 2018), which aligns with our study. Regardless of hammer mill screen size, the highest SD was achieved by cooling at 210 rpm screw speed across all feed materials. The effect was strongest in maize, likely because its continuous starch melt developed higher viscoelastic resistance when cooled, enhancing compaction during the die transition. Also, reduced expansion due to limited moisture evaporation may have contributed to the higher pellet density. In fiber-rich feed materials like barley, melt continuity may be disrupted, limiting compaction. In SBM-based treatments, since expansion was already limited without cooling, the additional effect of cooling on SD was minimal. Nevertheless, our study showed that combining 210 rpm screw speed with cooling the extruder’s last section enabled optimal pellet density for rumen escape (SD > 1.05 g/mL) in all feed materials except SBM.

In general, SDrf closely mirrored SD across processing conditions due to their strong correlation. However, maize processed with a 2 mm hammer mill screen and 210 rpm screw speed exhibited notably higher SDrf values than other feed materials, particularly barley. This suggests potential differences in internal pellet microstructure. Previous studies have shown that feed composition and processing variables can influence pellet microstructure (Zambrano et al., 2022; Ge et al., 2023), potentially affecting water absorption during rumen fluid incubation and contributing to variation in SDrf.

While appropriate SD is essential for achieving rumen escape or floating behavior, pellet stability in ruminal fluid remains crucial for functionality. Across all treatments, only barley and maize produced pellets with optimal FSI values (> 800 g/kg DM) after 90 min of rumen fluid incubation, and their FSI was significantly higher than that of SBM and its mixtures. Like density, pellet stability is influenced by both ingredient composition and processing conditions (Ma et al., 2021; Wang et al., 2021; Cheng et al., 2023). Key compositional factors include variations in starch, protein, fiber, and fat content as well as differences in starch properties (e.g., granule size, amylose:amylopectin ratio, amylose-lipid complex formation, and retrogradation ability) and protein characteristics (Shewry and Halford, 2002; Zhu, 2017), all of which affect the structural integrity of the extrudate (Zhang et al., 2014; Ge et al., 2023). Consistent with our findings, previous studies have shown enhanced pellet stability with increased starch content (Kannadhason et al., 2007; Larsen and Raun, 2018; Liu et al., 2021), a relationship linked to increased starch gelatinization and improved inter-particles binding (Rolfe et al., 2000; Welker et al., 2018; Cheng et al., 2022).

Surprisingly, barley exhibited higher FSI than maize, despite its higher fiber content, which is generally associated with reduced water stability (Larsen and Raun, 2018). We suggest that the lower expansion and the formation of a ’plywood’-like structure, created by the entanglement and folding of insoluble fiber particles (Thomas et al., 1998) within a starch matrix, may have improved fluid stability in barley. This explanation is supported by the reduced FSI in maize at increased screw speed, where enhanced expansion led to greater pellet porosity, allowing rumen fluid to penetrate and disintegrate the pellet. In contrast, cooling improved FSI in maize by enhancing compaction, reinforcing inter-particle bonds. For SBM-based treatments, as discussed earlier, extrusion-cooking conditions were not optimal. These suboptimal conditions likely hindered the formation of stable pellet structures and also explain the further reduction in FSI observed in M+SBM when cooling was applied to the extruder’s last section.

Overall, the extrusion conditions examined in this study produced pellets with sufficient fluid stability and a range of densities—from floating to high-density—likely to influence rumen escape in barley and maize, but not in SBM and its mixtures. Although the exact reasons for the behavior of SBM and its mixtures remain unclear, we speculate that future optimization may require increased thermal and mechanical energy. Simply increasing screw speed may not be effective, as it shortens residence time and can limit both starch and protein transformations (Lin et al., 1997). Alternate extrusion settings like a modified screw configuration with enhanced kneading and mixing, direct steam injection into extruder, or heating of extruder barrel wall can be explored. Moreover, as some SBM-based mixtures produced pellets with sufficient SD for rumen escape, their FSI might be improved using binders even under current processing conditions. Additionally, although feed moisture was held constant to isolate the effects of screw speed and cooling, its role in modulating melt viscosity, starch gelatinization, and matrix formation is well-documented (Draganovic et al., 2011; Fallahi et al., 2013) and should be considered in future optimization efforts, particularly for high-protein feed formulations.

4.3. Radial expansion and bulk density as proxies for SD and FSI

During feed processing, RE and BD are relatively easy to measure, so we examined their correlations with SD and FSI to determine if they could serve as proxies. Overall, RE and BD showed good correlations with SD and, to some extent, with FSI, although these relationships depended on the feed material and the range of densities measured. For instance, to achieve the minimum SD threshold of 1.05 g/mL required for enhanced rumen escape, we focused on pellets with SD values between 1.0 and 1.1 g/mL, which had corresponding BD values from 626 to 709 g/L. Within this range, no significant correlation between BD and SD was found, suggesting that BD alone may not predict whether the pellets reach the necessary SD threshold. If BD is to be used as a proxy, the pellets would likely need a BD above 710 g/L; nevertheless, we recommend direct SD measurements to confirm suitability for rumen escape. In addition, BD appears useful for categorizing feed pellets with respect to sinking/floating behavior. Our findings indicate that although pellets with a BD of 469 g/L exhibited floating behavior, we recommend using a threshold of below 430 g/L for floating pellets to ensure clear differentiation under rumen conditions. Similarly, pellets with a BD between 500 and 540 g/L should be considered slow sinking, 600–740 g/L as fast sinking, and above 740 g/L as very fast sinking. In contrast, no clear correlation was found between FSI and either RE or BD, indicating that these parameters cannot reliably substitute for direct FSI measurements. Hence, while RE and BD offer useful insights and may serve as proxies for certain pellet attributes, direct measurements remain essential for accurately assessing key quality metrics like SD and FSI.

5. Conclusion

Our study demonstrates that extruder processing can be optimized to produce feed pellets with tailored density and rumen fluid stability, both of which are critical for modulating rumen passage and degradation. Beyond the inherent effects of feed composition, the level of cooling in the extruder’s last section and extruder screw speed emerged as the most critical processing factors. For cereal grains, applying cooling at a screw speed of 210 rpm yielded pellets with optimal density and fluid stability for rumen escape. Without cooling, floating pellets were achieved at a minimum screw speed of 210 rpm for maize and 300 rpm for barley. In contrast, pellets produced from SBM or its mixtures with cereal grains did not meet the desired criteria under the tested conditions, indicating a need for alternative processing strategies. Furthermore, our findings show that measurements of bulk density and radial expansion cannot reliably substitute for direct assessments of specific density and fluid stability. Importantly, as the effects were evaluated using in vitro conditions, in vivo validation is essential to confirm the practical relevance of the findings.

This article was originally published in Animal Feed Science and Technology 329 (2025) 116482. https://doi.org/10.1016/j.anifeedsci.2025.116482. This is an Open Access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).