Impact of Process and Machine Parameters in the Charging Section on the Triboelectric Separation of Wheat Flour in a Vertical Separator

Abstract

Triboelectric separation has recently been investigated as a novel process for dry enrichment and separation of protein of various crops like wheat flour. The triboelectric effect allows for the separation of starch and protein particles in an electric field based on their different charging behavior despite having a similar density and size distribution. Particles are triboelectrically charged in a charging section before being separated in an electric field based on their polarity. While the charging section is crucial, the influence of process parameters remains largely unexplored. Thus, the influence of the charging sections’ dimensions and the particle concentration as process key parameters was investigated experimentally. Varying the length (0, 105, and 210 mm) showed that the protein shift increases with the length (max. 0.53%) during separation. Varying the diameter (6, 8, and 10 mm) influenced the charging behavior, resulting in an increase in protein accumulation on the negative electrode as the diameter decreased. Varying the mass flow of flour (40, 80, 160, and 320 g·h⁻¹) also affected the separability, leading to a maximum protein shift of 0.61%. Based on the observed results, it is hypothesized that the electrostatic agglomeration behavior of oppositely charged particles is directly affected by alterations in machine parameters. These agglomerates have a charge-to-mass ratio that is too low for separation in the electric field.

1. Introduction

In the 2024/25 harvest year, over 794 million tons of wheat was grown and processed worldwide by the leading industrial nations alone [1]. A large proportion of this amount is used for the production of wheat flour, which is mainly processed into food by bakeries and other manufacturers. The main component of wheat flour is starch, which makes up approx. 75–80% of the dry matter, and protein, which accounts for approx. 10–15% of the dry matter [2]. The protein content of wheat flour correlates directly with the resulting baked goods’ volume [3,4]. The protein content of wheat flour is therefore a decisive quality factor and the basis for price determination [5,6]. The composition of the wheat, as well as the protein content, is subject to regional and seasonal fluctuations as well as climate change [7]. As a result of recently arising stricter national regulations over the use of fertilizer, the produced flour already shows a decrease in protein content, as less nitrogen is incorporated into the soil, and therefore less protein can be formed in the wheat grain [8]. In addition, inconsistent protein contents in the milling industry must be balanced to ensure consistent output quality. For this reason, it is important to be able to adjust the protein content of wheat flour using simple separation processes during the milling process.

Existing methods include mixing different flour batches and adjusting the protein content by subsequent air classification after milling. Here, flour is further comminuted to break the grain’s structure into smaller particles, partially separating protein and starch [9]. The mostly smaller protein particles can then be separated by air classification based on the particle size, which allows protein enrichment [10]. However, this method is limited in practice due to high costs and low efficiency, as starch and protein particles have similar particle sizes [11].

The introduction of triboelectric separation during the milling process represents a novel method for inline adjustment of protein content in wheat flour, offering a new approach to improve the precision and the standardization of protein content. The separation characteristic is the charge of the particles instead of the size or the density. Here, the particles in the flour are electrostatically charged by the triboelectric effect, whereby starch particles usually become negatively charged and protein particles usually become positively charged during contact between each other [12,13,14]. The triboelectric effect occurs when two surfaces come into contact and then separate, transferring charge between them. A common example of triboelectric charging is rubbing a balloon against hair, which causes the balloon and the hair to become oppositely charged. The contact can take place through a collision or through a sliding movement of surfaces. During this contact, charge is transferred between the surfaces, which then remains on the surface [15,16]. This effect can be used to separate powders, as particles with opposite charges will move in different directions when placed in an electric field [17].

The actual separation process consists of particles being attracted to or deflected in the direction of the oppositely charged electrode due to their charge, whereby the electric field strength, mass of the particles, and their charge, as well as the residence time, are decisive influencing factors [14,17]. In addition to the separation setup, the charging section in which the particles are charged plays a central role. Different setups are used, including tubes, chutes, and rotating systems [18,19,20]. In rotating systems, like cyclones, particles are charged mainly through collisions between particles [21], while in chutes, charging occurs primarily through particle–wall interactions [22]. In tubes, both particle–wall and particle–particle collisions contribute to charging [19,23].

The triboelectric separation process is already used for mineral processing and plastic recycling on an industrial scale [24]. However, triboelectric separation of fine organic powders is still in its infancy and has not yet been used on an industrial scale. Previous studies of bean and soybean powders have shown that triboelectric separation can enrich protein content by over 25% [25,26]. Although it is known that the separation performance is highly product-dependent, to date, only limited data are available on the enrichment of protein in wheat flour. For inline standardization during milling, a small shift in the protein content of 1–2% would be sufficient. However, the influence of the process parameters on triboelectric separation of wheat flour has yet to be investigated. In the triboelectric separation of lupins, researchers have identified the design and conditions in the charging section as the most important factor influencing the triboelectric separation result [12,26].

Therefore, the influence of various process and machine parameters of the charging section on the triboelectric separation of wheat flour is examined. The main aim is to find out how changes in the length or diameter of the charging section affect the separation property. In addition, we aim to determine the influence of the particle concentration in the charging section on the separation properties.

2. Materials and Methods

2.1. Materials and Equipment

In all experiments, a single-variety wheat flour of Altdorfer Mühle GmbH (Altdorf, Germany), type 550, variety Asory, which was milled on 23 September 2021, was used with a protein content of approx. 12% (db).

2.2. Methods

2.2.1. Structure and Function of the Separator

The separator which is described in detail in [14] was used in the experiments and is shown in Figure 1. The separator setup was modified to fit the requirements of the experiments for wheat flour.

The powder is fed into the funnel through a vibrating conveyor (24.002, Fritsch GmbH, Idar-Oberstein, Germany) at the mass flow rates listed in

Table 1. Representation of the parameters: length of charging section [mm], diameter of charging section [mm], mass flow of flour [g·h⁻¹], residence time [ms], velocity [m·s⁻¹], and Reynolds number for the different tests (length and diameter of the charging section and mass flow of flour). Bold numbers are the parameters that were varied in the respective experiment.

	Length of Charging Section [mm]	Diameter of Charging Section [mm]	Mass Flow of Flour [g·h−1]	Residence Time [ms]	Velocity [m·s−1]	Reynolds Number
Test length	0	8	80	0	11.05	5849.77
	105	8	80	9.5	11.05	5849.77
	210	8	80	19	11.05	5849.77
Test diameter	210	6	80	10.7	19.65	7799.69
	210	8	80	19	11.05	5849.77
	210	10	80	29.7	7.07	4679.82
Test mass flow	210	8	40	19	11.05	5849.77
	210	8	80	19	11.05	5849.77
	210	8	160	19	11.05	5849.77
	210	8	320	19	11.05	5849.77

below. A grid is installed at the conveyor outlet to prevent the passage of large powder lumps. Dry air with a volume flow of 2 m³·h⁻¹ is fed into the funnel through a 6 mm diameter stainless-steel pipe coaxial to the charging tube to disperse the powder. The powder is dispersed in the funnel by the air flow and conveyed to the charging section below. In the charging section, which is made of a PVC tube, powder particles are triboelectrically charged through particle–wall and/or particle–particle contacts. The inlet represents the transition between the charging section and the separating section. The voltage source (HCE7-12500, FuG Elektronik GmbH, Schechen, Germany) is connected to the positive electrode (+) and induces a voltage of 12.5 kV. The negative electrode (−) is grounded via the voltage source. The electrodes are designed as parallel aluminum plates with a distance of 20 mm leading to an electrical field strength of 625 kV·m⁻¹. The electrodes are covered with PVC belts (1.5 mm thickness), which are driven by electric motors (919D Series, MFA Como Drills Ltd., Felderland, UK) using pulleys above and below the electrode and a belt connection coupled to the lower pulley. With the help of the pulleys, the belts rotate, and the deposited powder can be continuously removed by brushes, which are located at the back of the electrodes. The outlet finally separates the powder into four fractions. The external fractions (PE+ and NE−) contain the powder that reaches the electrodes. The powder that is not deflected sufficiently to reach the electrodes is collected at the bottom of the gap and is separated into two fractions, corresponding to the left (positive cup) and right (negative cup) sections of the gap (PC and NC).

2.2.2. Experimental Design

The length and diameter of the charging section were varied in order to investigate their impact on the flow characteristics and the target variable, protein shift ${∆\beta }_i$ [wt%] (weight percentage) (see Equation (6)). Variables such as charging section material (PVC), flow rate of dry air (2 m³·h⁻¹), and belt speed (0.01 m·s⁻¹) were kept constant to ensure comparability of the results. In addition, different mass flow rates of flour were examined in order to evaluate the influence of the particle–gas ratio on the protein shift ${∆\beta }_i$ [wt%].

To determine the effect of the residence time, tests were conducted with charging section lengths of 105 mm and 210 mm, as well as a test without a charging section (see Table 1). The diameter was kept constant at 8 mm. The mass flow rate was kept constant at 80 g·h⁻¹. Shortening the residence time lowers the total number of particle–particle and particle–wall interactions.

To evaluate the impact of diameter and thus the Reynolds number and residence time, charging sections with diameters of 6 mm, 8 mm, and 10 mm were tested (see Table 1). The length of the charging section was kept constant at 210 mm. The mass flow rate was kept constant at 80 g·h⁻¹.

In addition, the mass flow rate was varied by setting the values to 40, 80, 160, and 320 g·h⁻¹ (see Table 1). The fixed parameters of the charging section were a length of 210 mm and a diameter of 8 mm.

In order to determine the influence of the individual parameter changes, we decided to vary only one parameter at a time. For each variation, at least a double determination was carried out.

Table 2. Representation of the key parameters which were varied or kept constant for the tests: length of the charging section, diameter of the charging section, and mass flow of flour.

	Length of Charging Section	Diameter of Charging Section	Mass Flow of Flour
Test length	Varied	Constant	Constant
Test diameter	Constant	Varied	Constant
Test mass flow	Constant	Constant	Varied

shows the key parameters which were varied or kept constant in the experiments.

2.2.3. Moisture Content Determination

The moisture content had to be determined as measured protein content refers to the dry weight. The moisture content of the samples was determined using the drying oven method. Glass crucibles with lids were first dried for 30 min at 130 °C then cooled to room temperature in a desiccator for 30 min. Then, approx. 1 g of sample was weighed into each glass crucible with a precision balance (TLE204, Mettler Toledo, Columbus, OH, USA), which has a specified uncertainty of ±0.1 mg. The samples were dried at 130 °C for 2 h in a drying oven. At the end of the drying period, the jars were resealed with their lids to prevent moisture absorption and then cooled to room temperature in a desiccator for 30 min. The moisture content $x_{w,i}$ [wt%] (weight percentage) was calculated from the mass of the wet sample $m_{f,i}$ [g] and the dry sample $m_{t,i}$ [g] using Equation (1):

$$x_{w,i}={\displaystyle \frac{m_{f,i}-m_{t,i}}{m_{f,i}}}$$

(1)

A double determination was carried out for each sample, and the mean value of the two measurements was calculated.

2.2.4. Protein Content Determination

The protein content of the samples was determined using the Dumas method [27]. The organic elemental analyzer (vario MAX, Elementar Analysesysteme GmbH, Langenselbold, Germany) was used for the measurement (Method: Starch Powder). From the calculated absolute element content $a$ [g] and the sample weight $w$ [g], the element concentration $c$ [wt%] of the sample can be calculated using the daily factor $f$ according to Equation (2) (direct adaptation from the operating instructions of the organic elemental analyzer):

$$c={\displaystyle \frac{a·f}{w}}$$

(2)

The protein concentration ${\beta }_{w,i}$ [wt%] was determined from the resulting element concentration of nitrogen c, using a specific protein factor $F$, which for wheat flour is 5.83 [28]. The protein concentration ${\beta }_{w,i}$ [wt%] is defined by Equation (3):

$${\beta }_{w,i}=c·F$$

(3)

During triboelectric separation, the flour may dry due to the use of dry air for dispersion. This effect must be considered in the protein content calculations. The dry protein content ${\beta }_{s,i}$ [wt%] of a sample is calculated using the moisture content $x_{w,i}$ [wt%] and the protein concentration of the wet sample ${\beta }_{w,i}$ [wt%] to ensure comparability between the samples and the starting material. The dry protein content ${\beta }_{s,i}$ [wt%] is defined by Equation (4):

$${\beta }_{s,i}={\displaystyle \frac{{\beta }_{w,i}}{1-x_{w,i}}}$$

(4)

2.2.5. Particle Size Distribution

The particle size distribution was determined using laser diffractometer (Helos R-Series KR, Sympatec, Clausthal, Germany). For this purpose, approx. 3–5 g of the powder to be examined was analyzed. Lenses of the type R3 and R5 were used in combination in each measurement. The powder was first analyzed with the R3 lens, which covers a range of 0.5 to 175 µm, followed by the R5 lens, which covers a range of 4.5 to 875 µm. The measurement data were then combined. This combination covers a range of 0.5 to 875 µm.

2.2.6. Parameters for Characterization of Separation

To characterize the triboelectric separation process, the key values to characterize the separation are defined below. Since water content changes during separation, all values are based on dry weight. During the process, the added flour (standard or feed material, index “₀”) is separated into several fractions (index “_i”: two electrode fractions (PE+ and NE−) and two fractions between the electrodes (PC and NC) (see Figure 1). The mass yield, α [wt%], of a fraction i is defined by the mass of the standard material, $m_0$ [g], and the mass of the fraction, $m_i$ [g], as given in Equation (5):

$${\alpha }_i={\displaystyle \frac{m_i}{m_0}}$$

(5)

To estimate and evaluate the results of triboelectric separation, the protein shift, ${∆\beta }_i$ [wt%], is calculated using the dry protein content of the sample, ${\beta }_{s,i}$ [wt%], and the dry protein content of the standard (feed material), ${\beta }_{s,0}$ [wt%], as given in Equation (6):

$${∆\beta }_i={\beta }_{s,i}-{\beta }_{s,0}$$

(6)

2.2.7. Statistical Analysis

Two-way analysis of variance (ANOVA) with post hoc Bonferroni tests was performed to evaluate the significance of differences between groups. Group differences were deemed significant at p < 0.05. OriginPro 2023 software (OriginLab Corp., Northampton, MA, USA) was used for statistical analysis.

3. Results and Discussion

3.1. Influence of Length of Charging Section

The mass yield ${\alpha }_i$ as well as the protein shift ∆${\beta }_i$ of the experiments with different charging section lengths are shown in Figure 2.

Figure 2A shows that the mass is distributed across all fractions, with slightly more flour collected in the cups (PC and NC) than on the electrodes for all charging section lengths. No clear correlation between mass distribution and section length is observed.

Figure 2B indicates that for all configurations, the protein content was reduced with regard to the flour in the feed at the positive electrode. In the positive cup (PC), the flour was enriched in protein for all charging section lengths. Enrichment was also observed in the negative cup (NC), with slightly higher enrichment achieved as the charging section length increased. The protein shift at the negative electrode showed a mixed picture. In the experiment without a charging section, the protein content was reduced, while a greater reduction was observed with a charging section length of 105 mm. However, the protein content remained nearly unchanged with a 210 mm charging section.

As the length of the charging section increases, there is a slight increase in the accumulation of protein in the negative cup (except for 105 mm). This may be due to the stronger charging of the particles, resulting from the longer residence time in the charging section and, consequently, a greater number of particle–wall interactions. Similar findings were reported by Wang et al., who demonstrated that the specific charge of various PS particles increased as the tube length was extended from 125 to 225 mm. The authors attributed this increase in charge to the higher number of particle–wall collisions along the tube [29]. Similar findings were also reported by Xing et al., where a longer charging tube also resulted in a higher protein shift for soy flour. However, in their study, a spiral charging section was used, which likely influenced particle charging differently due to increased particle–wall interactions [26]. Matsusaka et al. also demonstrated that particle charge increases with the length of the charging section, using aluminum charging pipes in their study [30]. This is further supported by the particle size distribution (see

Table 3. Representation of the median values of the particle size distributions x₅₀ [µm] for the different charging section lengths 6, 8, and 10 mm at a gas volume flow of 2 m³·h⁻¹.

x50 [µm]	PE (+)	PC	NC	NE (−)
0 mm	53.49	93.92	95.81	55.65
105 mm	53.60	86.87	91.47	54.32
210 mm	54.54	84.05	90.23	62.59

), which shows that in the 210 mm experiment, larger particles were sufficiently charged and attracted to the electrodes, whereas in the 0 mm test, smaller particles were detected.

However, protein content was reduced at the negative electrode for nearly all charging section lengths (except for 210 mm). This is likely due to insufficient charging of the protein fraction, resulting in a low charge-to-mass ratio that prevents deflection onto the electrodes (see Figure 1). This suggests that the protein is either embedded in larger agglomerates that fall into the cup or that the charge is too low for separation. Here, only with the longest charging section is the protein not depleted. In the negative cup (NC), protein enrichment increases slightly with the length, likely due to more particle collisions and longer residence times promoting agglomeration. Interestingly, even without a charging section, particles carry enough charge for protein shift, likely generated during particle–conveyor contact or as they fall into the funnel. Yang et al. already demonstrated that vibrating conveyors can effectively triboelectrically charge pulse particles [31].

As residence time in the charging section increases, the risk of agglomeration of oppositely charged particles grows, as the particles are dispersed closely together, and the electric field between them becomes stronger. Only approx. 12% (db) protein is present in the initial flour. An agglomeration of oppositely charged starch and protein particles, assuming similar charge tendencies, leads to charge equalization (neutralization) and results in a higher protein content in the newly formed agglomerate compared to the original flour. It can also be assumed that the charge of the protein fraction is higher when the length is increased [26,29,32], which makes agglomerate formation more likely due to stronger attraction of the oppositely charged particles. Starch particles in flour are typically larger and have a greater surface area than protein particles. This increases the likelihood that a starch–protein agglomerate, once formed, could become negatively charged again during subsequent collisions, binding additional protein particles.

One key finding from varying the charging section length and thus particle residence time is that a longer residence time results in a stronger charge of the protein fraction, leading to a slightly higher protein content in the negative fractions (NE and NC).

3.2. Influence of Charging Section Diameter

The mass yield ${\alpha }_i$ as well as the protein shift ∆${\beta }_i$ of the experiments with different charging section diameters are shown in Figure 3.

Figure 3A shows that the distribution of flour into different fractions is influenced by the charging section diameter. As the diameter increases, less flour is collected on the electrodes and more in the cups. Specifically, for the 6 mm tube, the mass yield in the cups is lower than on the electrodes, while for the 8 mm tube, the yield in the cups exceeds that on the electrodes, with this effect being most pronounced for the 10 mm tube. Additionally, mass loss due to increased flour swirling in the funnel increases as the charging section diameter decreases.

Figure 3B shows that the protein shift across the different charging section diameter configurations reveals a trend in the individual fractions. Protein is depleted at the positive electrode in all experiments, with a slightly stronger depletion observed as the tube diameter increases. Protein is enriched in both the positive and negative cups for all diameters, with a slight increase in protein content as the diameter increases. The negative cup with a 10 mm charging section diameter showed the highest protein enrichment, at 0.59%. The protein shift at the negative electrode appears to depend on the diameter. A 6 mm diameter resulted in a 0.48% enrichment, which decreases with larger diameters and becomes slightly negative for a diameter of 10 mm.

As the diameter decreases, a steady increase in the mass yield at the electrodes and a decrease in the cups can be observed. This suggests that a smaller diameter leads to a stronger charge of the particles, which also seems plausible due to the increased air velocity in the charging section and the associated increased Reynolds number. Similar findings were reported by Wang et al., who observed that particle charging increases with higher turbulence, which they induced by increasing the gas flow rate [13].

The protein shift at the negative electrode shows that reducing the diameter increases protein fraction charging and improves accumulation, with protein enrichment consistently increasing as the diameter decreases. This is further supported by the particle size distribution (see

Table 4. Representation of the median values of the particle size distributions x₅₀ [µm] for the different charging section diameters 6, 8, and 10 mm at a gas volume flow of 2 m³·h⁻¹.

x50 [µm]	PE (+)	PC	NC	NE (−)
6 mm	70.74	100.47	102.69	67.82
8 mm	54.54	84.05	90.23	62.59
10 mm	52.1	84.3	94.61	57.83

), which shows that in the 6 mm experiment, larger particles were sufficiently charged and attracted to the electrodes, whereas in the 10 mm test, mostly smaller particles were detected. Reducing the charging section diameter not only enhances particle charging but also enables the charging of larger particles, increasing the charge-to-mass ratio.

In contrast to this, the highest protein shift was observed in the cups. As the diameter of the charging section increased, slightly more protein was enriched in each cup. A larger diameter likely promotes greater agglomerate formation due to enhanced interactions between oppositely charged particles. This can be explained by the increasing residence time with increasing diameter and decreasing Reynolds number. With a 10 mm diameter, the residence time is nearly three times longer than with a 6 mm diameter (see Table 1). The longer residence time, along with the lower velocities and reduced Reynolds number and thus turbulence intensity, increases the likelihood of agglomerate formation between oppositely charged particles. Wang et al. also demonstrated that the agglomeration of oppositely charged gluten–starch particles can be minimized by using high gas velocity, which increases turbulence [13].

High turbulence intensity can break up previously formed agglomerates in the charging section, which could explain the smaller accumulation of protein in the cup for the charging section with a 6 mm tube diameter. In addition, the probability of agglomerate formation in the separation chamber is also greater due to a lower entry velocity (at a higher charging section diameter) in the separation section as both sections are connected to each other and consequently have the same volume flow (see Figure 1). Therefore, the residence time in the separation section is longer for the highest charging section diameter. The entry velocity into the separation section, which describes the speed of the air and the particles after the charging section, is almost three times as high for a charging section with a diameter of 6 mm as for one with a diameter of 10 mm (see Table 1). However, it should be noted that the highest positive protein shift was achieved by using the charging section with a diameter of 10 mm. This raises the question of whether the formation of agglomerates during the triboelectric separation of flour should be aimed at or prevented with the goal of maximizing enrichment.

The main finding from varying the diameter is that reducing the charging section diameter increases particle charge and protein accumulation at the negative electrode. This is likely due to a decrease in agglomerate formation, with increased turbulent shear stress breaking up existing agglomerates. As a result, protein accumulation in the cup is lower, and the maximum protein accumulation is reduced.

3.3. Influence of the Feed Mass Flow Rate

The mass yield ${\alpha }_i$ as well as the protein shift ∆${\beta }_i$ of the experiments with different mass flows of flour are shown in Figure 4.

Figure 4A shows that, for all mass flow rates examined, more mass was collected in the cups than on the electrodes, with slightly more flour collected in the negative cup than the positive cup. The proportion of flour mass collected in the cups increases slightly with higher mass flow rates.

Figure 4B shows that at the positive electrode (PE+), protein depletion increased slightly with higher mass flow rates, except in the 80 g·h⁻¹ experiment. Protein enrichment was observed in the cups across all experiments, regardless of the mass flow rate. The highest protein enrichment of 0.61% was achieved with a mass flow rate of 160 g·h⁻¹ in the positive cup (PC). At the negative electrode, the protein shift shows a significant dependence on mass flow rate: lower mass flow results in higher protein content at the negative electrode.

The dependence of protein shift at the negative electrode on the feed mass flow rate suggests that the protein fraction is more strongly charged at lower mass flows. This can be attributed to better dispersion and a longer mean distance between particles at lower concentrations, which results from the reduced mass flow. Matsusaka has already shown that a higher particle charge can be achieved by reducing the mass flow [33]. According to his study, the particle–wall impacts in the charging section increase with a reduction in the mass flow. This effect was also highlighted by Wang et al., who found that the specific charge of a gluten–starch mixture decreased significantly with increasing particle concentration. According to their study, this decrease in charge is attributed to a reduced likelihood of particle–wall collisions as mass flow increases [13].

This contrasts with the contact frequency of the particles, which increases with increasing mass flow rate due to a smaller mean distance between particles. In practice, however, it is possible that lumps of flour or large agglomerates do not disperse at all or only slightly when the mass flow rate is high, thus worsening the charging behavior of the individual particles due to the resulting smaller total surface area of the particles. This effect is also evident in the particle size distribution of the different fractions (see

Table 5. Representation of the median values of the particle size distributions x₅₀ [µm] for the different mass flows 40, 80,160, and 320 g·h⁻¹ at a gas volume flow of 2 m³·h⁻¹.

x50 [µm]	PE (+)	PC	NC	NE (−)
40 g·h−1	53.28	82.22	83.89	75.67
80 g·h−1	54.54	84.05	90.23	62.59
160 g·h−1	51.08	86	83.13	64.44
320 g·h−1	52.69	86.77	77.96	55.79

), where a smaller proportion of large particles is attracted to the negative electrode as mass flow increases. Consequently, the x₅₀ of particles at the negative electrode steadily decreases with increasing mass flow rate, except at 80 g·h⁻¹.

In addition, no significant trend of the protein shift in the cups as a function of the mass flow rate is notable. The shift is lowest at mass flows of 40 and 320 g·h⁻¹ and highest at 80 and 160 g·h⁻¹. The accumulation of protein in the cups could be explained by the formation of agglomerates from previously oppositely charged protein and starch particles. Agglomerate formation seems to be strongest at mass flows of 80 and 160 g·h⁻¹, which can be recognized by the higher protein enrichment. One possible explanation for this is the dependence of charge on powder dispersion [13]. At low mass flow rates, better dispersion allows more protein particles to become positively charged. However, the larger distance between particles in the gas flow reduces the attraction between them, lowering the probability of agglomeration despite the higher charge. At high mass flow rates, particle contact frequency increases in both the charging and separation sections, but poorer dispersion likely results in fewer protein particles being charged. As a result, agglomeration seems less likely in this case. At 80 and 160 g·h⁻¹, a sufficient particle density in the charging and separation section interacts with a sufficient charge of the protein fraction, which leads to a high amount of agglomerate formation and a resulting strong protein shift of up to 0.61%.

The main findings from varying the mass flow rates are that protein shift at the negative electrode increases with a decreasing particle–gas ratio, as shown by various authors, and that agglomeration formation slightly increases with better dispersion and stronger particle charging.

4. Conclusions

This study examined the impact of machine parameters such as the length and the diameter of the charging section of a vertical triboelectric separator and how they affect separation characteristics. In addition, the influence of particle concentration was assessed. The variation in length led to different residence times at a constant Reynolds number and thus turbulence intensity. It was found that a longer residence time led to a higher charge of the protein fraction, but also to a higher probability of agglomeration. The variation in diameter influenced both the residence time and the Reynolds number. It was found that a smaller diameter led to a higher charge of the protein fraction, as well as to a lower tendency to agglomerate, due to higher turbulence intensity. Finally, the charge of the protein fraction is higher for lower particle concentrations, as the powder dispersion is improved and particle–wall collisions are more frequent.

In all experiments, protein was enriched in the cups and depleted at the positive electrode. This is in line with the resulting charge regarding the order of protein and starch in the triboelectric series. On the other hand, the protein concentration at the negative electrode was strongly dependent on the process parameters, with a length of 210 mm, a diameter of 6 mm, and a mass flow of 40 g·h⁻¹ leading to the highest protein shifts. This dependency can be mainly explained by the formation of agglomerates and the charge-to-mass ratio of the formed agglomerates.

The observed enrichment in protein content is much lower than in other plants like pulses. The reason for this could be the different grain structure and the separability of protein and starch during grinding as well as different grinding methods. In pulses like navy beans, protein shifts of more than 20% can be achieved. However, since triboelectric separation of wheat flour could be used to standardize the protein content in line, an enrichment of ca. 1–2% may be sufficient to compensate for seasonal or geographical fluctuations and seems achievable with an integrated optimization of the process. In addition, this study focused solely on the effects of changes to the charging section, neglecting the separation section and flour dispersion, which could also impact separation results. This study can serve as a basis for the triboelectric enrichment of protein from other plant sources such as pulses.

The results gained from the variation of the process parameters as well as particle size distributions suggest that electrostatic agglomeration may play a major role. Electrostatic agglomerates of oppositely charged starch and protein particles may form, which strongly influence protein enrichment. These agglomerates are minimally deflected in the electric field in the separation zone due to a low charge-to-mass ratio. The electrostatic agglomeration behavior of the particles is affected by changes in key parameters, raising the question of whether targeted agglomeration should be pursued to maximize the protein shift. This will be further investigated in future studies.

Finally, it can be stated that triboelectric separation may be a useful and cost-effective way to standardize wheat flour with regard to the occurring fluctuation of protein content after milling in the future, which can also be implemented directly into the milling process as a continuous unit operation.