Optimised dry processing of protein concentrates from Australian pulses: A comparative study of faba bean, yellow pea and red lentil seed material

Milling and air classification settings were optimised for production of protein concentrates from Australian faba bean, yellow pea and red lentil seed material. Pulses were milled to flour of three progressively finer particle size distributions (D50 of 23–25, 16–18 and 13–14 μm) and air classified at classifier wheel speeds of 7080, 9600 and 10,200 rpm. Maximum protein concentration was reported for pulse flours (D50 of 13–14 μm) at 9600 rpm. Protein concentrations of 61.4, 58.1 and 61.0 g/100 g (db.), reflecting a fold increase in protein content of 1.9, 2.3 and 2.1, were reported for faba bean, yellow pea and red lentil, respectively. Protein, ash, fat and total dietary fibre contents were significantly higher in fine fractions (p < 0.05), compared with coarse fractions, resulting in protein concentrates with enhanced nutritional properties. Amino acid score (AAS) of protein concentrates highlighted deficiencies in sulphur‐containing amino acids, methionine and cysteine (MET + CYS), and tryptophan. Based on the lowest AAS (MET + CYS), protein concentrates were ranked highest for yellow pea (0.75), followed by faba bean (0.58) and red lentil (0.51). Phytochemical analysis demonstrated that bioactive constituents also co‐concentrated with protein (fine fraction), potentially leading to protein concentrates with enhanced health benefits. Shelf‐life assessment for the original flours and protein concentrates indicated the onset of rancidity after 3 months of storage. As fat content co‐concentrated with protein, the rancidity (%) scores were higher for protein concentrates compared with the original flours. This demonstrates the importance of developing effective treatments, suitable for dry processing, which can extend shelf‐life and stability of protein concentrate ingredients for domestic and export markets. The objective of this study was to increase the knowledge available on dry processing of protein concentrates from Australian pulses. The information generated from this study will look to inform future commercial scale processing operations.


| INTRODUCTION
Pulses are dry and edible seeds from legumes that are grown and consumed worldwide. Pulses are often grown in rotation with cereals and provide a range of agronomic benefits such as increased crop diversity; breaking of cereal disease and weed cycles; improved soil water retention; fixing of atmospheric nitrogen into the soil, thereby, reducing the need for nitrogen fertilisers; and lowering the carbon footprint (Jensen et al., 2010;Liu et al., 2016;Stagnari et al., 2017). Australian pulse production has grown rapidly over the last few decades and is driven by plant breeding initiatives tailored to deliver new varieties with higher yields, improved disease resistance, wider adaption and seed quality attributes preferred for export markets, such as seed size and colour (Siddique & Sykes, 1997). Australian pulse production averages 2.2 million metric tonnes annually, with the major pulse crops comprising lupins, chickpeas, faba beans, field peas, lentils and mungbeans (Australian Export Grains Innovation Centre [AEGIC], 2017). In Australia, average production of faba beans, yellow peas and red lentils, over the 5-year period from 2016-17 to 2020-21, was 391,000, 279,000 and 578,000 t, respectively (Australian Bureau of Agricultural and Resource Economics and Sciences [ABARES, 2022], https:// www.agriculture.gov.au/abares). The adoption of strict industry standards across the entire supply chain ensures consistent supply of high-quality pulses for human consumption, with majority of production destined for export markets, including Egypt, Saudi Arabia, China, India, Bangladesh, Sri Lanka and Pakistan (AEGIC, 2017; Pulse Australia, 2022, http://www.pulseaus.com.au).
Pulses are high in protein and provide a good source of carbohydrate, dietary fibre, vitamins, minerals and phytochemicals (Boye et al., 2010;Day, 2013;Hall et al., 2017;vaz Patto et al., 2015). Pulses are processed for human consumption in the form of whole or split seed, or milled to flour, for use in many traditional or snack-based foods. In addition, wet and dry processing methods enable fractionation of pulses into their major constituents (protein, starch and fibre), for a broader range of end-use applications, including plant-based meat and dairy alternatives, bakery products and nutritional supplements (Asgar et al., 2010;Boye et al., 2010;Byanju & Lamsal, 2021;Sozer et al., 2017;Vogelsang-O'Dwyer et al., 2021). Historically, commercial sources of plant protein have been dominated by wheat, soybean and pea. The rapidly growing global demand for plant protein provides an opportunity to leverage a wider range of pulses, for processing into value-added protein ingredients with enhanced nutritional and functional properties.
Dry processing is energy and water efficient, providing an economic and sustainable alternative to wet processing methods (Assatory et al., 2019;Berghout et al., 2015;Schutyser et al., 2015).
Dry processing includes milling and air classification technology for production of protein and starch concentrates that retain native functionality. Impact classifier milling is preferable for producing flour of fine particle and low starch damage, which is suitable for air classification. Optimal milling conditions are required to disrupt cotyledon cell constituents and detach small protein bodies (<10 μm) from larger starch granules, ranging from 20 to 40 μm in size (Schutyser & van der Goot, 2011). The degree of detachment during milling is influenced by a range of factors, including milling intensity, seed moisture, seed hardness, starch granule size and chemical composition, such as crude fibre and water-insoluble cell wall content (Assatory et al., 2019;Tyler, 1984). Air classification separates flour into coarse (starch concentrate) and fine (protein concentrate) fractions, based on physical differences in particle size, shape and density, in an airflow. The air classification process can be manipulated by adjusting classifier wheel speed and airflow settings. Other nutritional components will coconcentrate with protein in the fine fraction. This is partly attributed to the intrinsic composition of protein bodies, containing protein contents of 70-88 g/100 g (db.), and minor constituents such as lipids and sugars (Plant & Moore, 1983;Schutyser et al., 2015).
Milling and air classification of pulses were first reported by Youngs (1975) and subsequently by others, with much of this earlier work focused on peas (Reichert, 1982;Reichert & Youngs, 1978). Further application of this technology to a wider range of pulses has recently been reviewed and highlights protein concentration levels typically ranging from 40 to 65 g/100 g, depending on the type of pulse (Assatory et al., 2019;Fernando, 2021;Schutyser et al., 2015).
Further studies have focused on developing methods to optimise protein concentration from pulses.  reported a facile method to increase protein concentration from pea, bean, lentil and chickpea. In this study, optimal milling conditions for detaching protein bodies from starch granules were achieved when particle size distribution curves of the original pulse flour and isolated starch granules overlapped maximally. Subsequent air classification of pea, bean, chickpea and lentil flours achieved protein concentrations of 55.6, 52.8, 45.3 and 58.5 g/100 g (db.), respectively. A further study by Pelgrom, Wang, et al. (2015) investigated the effect of pretreatments to enhance protein concentration from pea and lupin, including varying seed moisture content levels, defatting, soaking and freezing cycles. Higher seed moisture contents and defatting led to increased protein concentration for lupin. Response surface methodology (RSM) was used to investigate the effect of two air classification variables (classifier wheel speed and airflow settings) on the composition, yield and anti-nutritional content of air-classified fractions from field peas (Wang & Maximiuk, 2019). The authors reported good agreement between experimental and predicted values obtained using the models developed. Further application of RSM was used to investigate the effect of three variables (classifier wheel speed, feed rate and airflow setting) on protein concentration and protein yield for mungbean (Zhu et al., 2020). However, one limitation of the research published in this space is that most studies were conducted at laboratory or pilot scale. Rempel et al. (2019) recently addressed this issue by reporting the industrial scale milling and air classification of commercially grown peas in Canada, which resulted in pea protein concentration levels ranging from 42 to 50 g/100 g (db.).
There is now considerable opportunity in Australia for commercial scale dry processing operations to form an important component of a broader domestic pulse protein processing industry. As the global demand for plant protein continues to rise, the efficient processing of Australia's major pulse crops into value-added protein ingredients will help to meet this demand. The Australian pulse industry can capitalise on this opportunity through the development of premium varieties, which will meet manufacturers' demands for protein ingredients with enhanced nutritional, functional, shelf-life and sensory attributes. As domestic manufacturing increases, demand for locally produced pulse protein ingredients will lead to greater investment in commercial processing operations. Australian growers will benefit from this increased demand, having the opportunity to expand their pulse production and supply new and more lucrative markets. The study aims to increase our technical knowledge for efficient dry processing of protein concentrates from Australian pulses, with an initial focus on faba beans, yellow peas and red lentils. Technical information generated on Australian pulses will look to inform domestic commercial scale processing operations.

| Starch isolation
Starch granules were isolated from pulse flours according to the method of Joshi et al. (2013), with slight modifications. Flour samples (80 g) were suspended in distilled water (1:10 w/v), and the pH was adjusted to 9.0 with 1 M NaOH/HCl. The suspension was mixed using a magnetic stirrer for 3 h at room temperature to solubilise the protein and facilitate the separation of starch. The mixture was centrifuged at 17,500 g for 15 min (Sorvall Lynx 4000 centrifuge) and the supernatant was discarded. The starch pellet was washed with 500 ml of distilled water and centrifuged again and the supernatant was discarded. This washing step was repeated a further three to four times as required, with careful removal of the greyish layer (insoluble protein) residing on top of the starch pellet. After washing, the starch pellet was redispersed in 250 ml of distilled water and passed sequentially through a 250 and 90 μm screen. The starch suspension was centrifuged at 5000 g for 15 min and the supernatant was discarded. The starch pellet was dried in an oven at 35 C for 24 h, then gently ground using a mortar and pestle and sieved using a 150 μm screen.
Starch granule isolates were stored at 4 C until further use.

| Particle analyses
Particle size distribution of pulse samples was measured in duplicate using a Saturn DigiSizer ® II High-Definition Digital Particle Size Analyser (Micromeritics, USA). Water was used as the dispersing medium and the suspensions were subject to sonication to aid dispersion. The volume-weighted particle size distribution was reported. The particle size distribution defines the relative amount, usually by volume (%), of particles present in a sample according to size (μm). Particle size distributions were characterised by D 10 , D 50 and D 90 values, representing 10%, 50% and 90% of volume-based cumulative size, respectively.
The median particle size (D 50 value) indicates that 50% of the volume is made up of particles smaller than this size and that 50% of the volume is made up of particles larger than this size. Starch granule morphology was studied using a Phenom XL scanning electron microscope (SEM). A thin layer of sample was directly applied to double-sided adhesive carbon tape on an aluminium stub and was coated with gold. Samples were examined using an accelerating voltage of 10 kV (3000-4000Â magnification).

| Impact classifier milling and air classification of pulses
The moisture content of faba bean, yellow pea and red lentil was 10.4, 11.1 and 7.3 g/100 g, respectively. Prior to impact classifier milling, the red lentil seed material was conditioned to a moisture content of $10.0 g/100 g, by adding the appropriate amount of water.
Pulses were milled to specification using a Hosokawa ACM10 impact classifier mill. Pulse flours of three progressively finer particle size distributions were produced, with D 50 values of 23-25, 16-18 and 13-14 μm. Subsequent air classification was conducted using a pilot scale air classifier (FW120 Model air classifier; Weifang ALPA Powder Technology & Equipment Company, China). Flour batches of 2 kg were air classified at a feed rate of $2 kg/h using a vibratory feeder.
Total flour recovery rates after air classification were generally >95% (w/w). Classifier wheel speed settings of 7080, 9600 and 10,200 rpm were used in this study. The adjustable air valve located at the bottom of the classification chamber was set at 25% open for all air classification trials. Airflow entering the bottom of the classification chamber aids in particle dispersion and forces particles up towards the classification zone. Air-classified coarse (starch concentrate) and fine (protein concentrate) fractions were weighed for mass balance and analysed for moisture and protein content. Mean values ± standard deviations (n = 2) from duplicate air classification trials were reported for protein content (g/100 g, db.) and protein yield (%). Protein yield refers to the percentage of total flour protein recovered in the fine fraction.

| Amino acid analysis
Amino acid composition of protein concentrates was determined as previously described (Skylas et al., 2017). The amino acid score (AAS) of protein concentrates was calculated by comparing the essential amino acid composition of protein concentrates to a standard universal reference pattern, which in this study was based on the amino acid requirements established for pre-school children aged 2-5 years old (Protein Quality Evaluation Report, FAO/WHO Expert Consultation, 1991).

| Phytochemical composition
All reagents used to determine the phytochemical composition of the original pulse flours and protein concentrates were of analytical grade.
Methanol, sodium hydroxide and sodium carbonate were purchased from Chem Supply, and all other reagents were purchased from Sigma-Aldrich Australia. The free phenolic compounds were extracted with 90% methanol, using the protocol reported by Johnson, Walsh, et al. (2021). An extraction mass of 0.5 g and total solvent volume of 14 ml were used. All extractions were performed in triplicate. The bound phenolic compounds were extracted from the sample pellet remaining after extraction of the free compounds, using alkaline hydrolysis. This procedure was modified from Yüksekkaya et al. (2021). Firstly, 7 ml of 2 M NaOH was added to the pellet, which was then shaken end over end for 60 min. After adjusting the pH of the extracts to 1.5-2 using concentrated (32%) HCl, the samples were centrifuged (1000 g for 10 min) and the supernatant was collected.
Following addition of a further 7 ml of 90% methanol to the pellet, samples were shaken end over end for 20 min and centrifuged and the supernatant was again collected. The combined supernatants for each sample were made up to 15 ml volume using 90% methanol. Previously reported methods were used to determine the total polyphenolic content (TPC) (Johnson, Mani & Naiker, 2021) and ferric reducing antioxidant potential (FRAP) (Johnson, Walsh, et al., 2021)

| Accelerated shelf-life testing
Accelerated shelf-life testing was conducted to investigate the onset of rancidity for pulse flours and protein concentrates. Samples (200 g) for each time interval were placed in individual clear snap-lock bags and stored at 30 C and 65% relative humidity, in an incubator with clear glass doors. One week of storage in these conditions is modelled to replicate 2 weeks of storage in ambient conditions ($22 C). Sensory assessment and laboratory testing of samples (FFA, a w and pH) were conducted at receival (initial receipt after air classification) and after 3 months of storage. For sensory assessment, two tablespoons of sample were placed in a cellophane bag for each panellist and incubated at 25 C for 1 h. Sensory assessment of samples was conducted in temperature-controlled neutral-coloured booths by six panellists, who were instructed to score the degree of perceived rancidity on 15 cm linear scales, which is similar to the method reported by Heiniö et al. (2002). This score was converted to a percentage value, in which the higher the score, the more rancid the sample. To assist in the sensory assessment of pulse flours and protein concentrates, the panellists were repeatedly exposed to rancid material, including rancid and non-rancid control samples. The panellists were then selected on their ability to consistently detect known rancid samples.

| Statistical analysis
Where replicate analyses were performed, the results are presented as mean ± 1 standard deviation. Statistical analysis was performed on the replicated data using R Studio running R 4.0.2 (R Core Team, 2020). As all data appeared parametrically distributed based on their histograms, one-way ANOVAs followed by post hoc Tukey's testing were used. A significance level of α = 0.05 was taken as statistically significant.

| Nutritional composition of pulses
The nutritional composition for a diverse range of pulses has been well documented in the literature (Hall et al., 2017;Vaz Patto et al., 2015). In this study, dehulled and split seed samples for faba bean, yellow pea and red lentil were milled to flour and analysed for nutritional parameters, including protein, ash, fat, total starch and TDF. The nutritional composition differed significantly between pulses for all nutritional parameters tested (p < 0.05). Protein content was 33.0, 25.7 and 28.5 g/100 g; ash was 3.2, 2.9 and 2.4 g/100 g; fat was 1.7, 1.2 and 1.5 g/100 g; total starch was 43.1, 54.8 and 48.9 g/100 g; and TDF was 8.7, 10.5 and 9.2 g/100 g, for faba bean, yellow pea and red lentil, respectively. There were considerable differences in protein content, being much higher for faba bean. The general trend for all three pulses was the higher the protein content, the lower the starch content. Ash content is a measure of mineral content, with pulses known to be good sources of potassium, magnesium, iron, manganese, zinc, copper, selenium and calcium (Hall et al., 2017). The fat content was relatively low, being <2 g/100 g for all three pulses, which is an important property for efficient air classification. Other pulses such as chickpea (2-7 g/100 g) and lupin (5-15 g/100 g) have much higher levels of fat (Hall et al., 2017), which can lead to poor flour dispersibility properties and less efficient air classification. Starch is the most abundant carbohydrate in pulses and was found to be highest for yellow pea, followed by red lentil and faba bean. Pulse starches have a lower glycaemic index compared with cereal starches, meaning they release glucose into the blood stream at a much slower rate and, therefore, can be used in the formulation of specialised diabetic food products (Vaz Patto et al., 2015). The composition, structure and functional properties of pulse starches have previously been reviewed (Hoover et al., 2010;Ren et al., 2021

| Impact classifier milling of pulses
Optimal milling conditions are required to produce flour with a fine particle size distribution, being suitable for air classification. To facilitate the milling process, starch granules were isolated from pulse flours and their respective size distributions were determined.
Starch granule size influences the efficiency of air classification, with larger starch granules having greater potential for separation from protein bodies .  Table 1, with the corresponding particle size distribution profiles shown in Figure 1. Increasing classifier wheel speed during milling led to a decrease in flour particle size, resulting in a significant decrease in moisture content between flours milled from the same pulse (p < 0.05). This is due to the longer residence time of finer particles in the mill chamber. There was a significant difference in protein content between the flours milled from red lentil, with protein content increasing as flour particle size decreased (p < 0.05).
Starch damage was relatively low for all flours, ranging from 1.3 to 2.4 g/100 g, being highest for red lentil and lowest for faba bean.
There was no significant difference in starch damage between flours of progressively finer particle size from the same pulse (p > 0.05), which is consistent with that previously reported by Pelgrom et al. (2013). However, it is important to avoid excessive milling conditions, leading to broken or fragmented starch granules of similar size to protein bodies, as this will have a negative impact on efficient separation of protein and starch during air classification. Milling of flour to a very fine particle size can also increase the surface area, promoting flour cohesiveness and poor dispersibility properties in an airflow. In Figure 1, the large narrow peak (23-25 μm) in the particle size distribution represents the starch granules, whereas the smaller and broader peak between 1 and 10 μm represents the detached protein bodies, or fragments of protein and cell wall material. The peak area between 1 and 10 μm was largest for flours milled to D 50 of 13-14 μm, resulting from a greater degree of milling intensity. This is also reflected in Table 1 with the increased volume of particles being of <10 μm in size ($40%).

| Air classification of pulse flours
3.3.1 | Preliminary trials to establish optimal settings for protein concentration Preliminary air classification trials were conducted to establish optimal classifier wheel speed and airflow settings for protein concentration (data not shown). Faba bean was selected for these preliminary trials fractions were weighed and analysed for moisture and protein content. The resulting mass balance for the fine fractions is reported in Table 2. The maximum protein concentration and protein yield for faba bean at 9600 rpm were achieved for flour milled to the finest particle size (D 50 of 13.5 μm). At 9600 rpm, protein content of 61.4 g/100 g and protein yield of 47.2% were significantly higher compared with faba bean flours milled to D 50 of 18.2 and 23.3 μm (p < 0.05). The general trend for faba bean flour at 9600 rpm was the finer the particle size, the greater the mass, and the higher the protein content and protein yield in the fine fraction. This is consistent with the increased volume of particles <10 μm for flours milled to a finer particle size (Table 1) Table 3. Protein concentration from yellow pea and red lentil flour was much higher (>55 g/100 g) compared with the previous trials reported in

| Particle size distribution and morphology of fine fractions
Particle size distributions were analysed for fine fractions (protein concentrates) produced from air classification of pulse flours (D 50 of T A B L E 2 Mass balance, protein content and protein yield of air-classified fine fractions from pulse flours milled to three different particle size distributions 13-14 μm) at 9600 rpm ( Table 2)  Coarse and fine fractions were produced from air classification of pulse flours (D 50 of 13-14 μm) using a classifier wheel speed of 9600 rpm.
in Table 2, whereas fractions from yellow pea and red lentil are from the optimised trials reported in Table 3. These fractions were selected based on fine fractions having the maximum level of protein concentration for each pulse. Resulting data in Table 4

| Amino acid composition and AAS of protein concentrates
The amino acid composition of the pulse protein concentrates is reported in Table 5. The proportion of essential amino acids, relative to total amino acids, is fairly conserved between the three protein concentrates, ranging from 37.6% to 39.1%, being highest for yellow   TMAC, total monomeric anthocyanin content; TPC, total polyphenolic content.
The quality of a food protein source relates to the amino acid composition, and the ability of that protein to be digested, absorbed and metabolised by the human body. Factors that can influence the quality of dietary proteins and the implications for pulses have previously been reviewed ). An initial basic measure of protein quality is the AAS, which is generated by comparing the essential amino acid composition of a protein source with a universal reference pattern (  Table 6 show all three pulse protein concentrates being deficient in MET + CYS and tryptophan (TRP), with faba bean also deficient in threonine (THR). The lowest AAS (MET + CYS) is used to compare the quality of different sources of protein. Based on this, protein concentrates were ranked highest for yellow pea (0.75), followed by faba bean (0.58) and red lentil (0.51). This is consistent with that reported by , in which the AAS for pea, faba bean and lentil protein concentrates was 0.72, 0.65 and 0.56, respectively.

| Phytochemical analysis of pulse flours and air-classified fractions
The antioxidant capacity, total phenolic content (TPC) and anthocyanin content from the free and bound phenolic extracts are reported in Table 7. Previous research reported in the literature has shown that phenolics and antioxidant compounds fractionate differently between air-classified coarse and fine fractions (Cammerata et al., 2021;Inglett & Chen, 2011). These compounds impart a range of potential health benefits to pulses and pulse-based products, including reduced inflammation (Milesi et al., 2022) and lowering the risk of developing cardiovascular disease (Rangel-Huerta et al., 2015) and diabetes (Charoensiddhi et al., 2022). However, there has been little attention given to the impact of air classification on these potential healthbenefitting compounds, with the primary focus being on production of protein and starch concentrates (Assatory et al., 2019;Fernando, 2021). Consequently, the antioxidant capacity, TPC and anthocyanin content were analysed for the original pulse flours, as well as the air-classified coarse (starch concentrate) and fine (protein concentrate) fractions. These analyses were conducted separately for free (methanol-soluble) phenolic extracts and the esterically bound phenolic extracts.
For the free phenolic extracts, one-way ANOVA testing demonstrated that the TPC was significantly higher in the fine fraction compared with the original flour for all pulses. This difference was greatest for faba bean (143% higher TPC in the fine fraction) and lowest for the red lentil (94% higher TPC). Furthermore, the TPC of the coarse fraction from yellow pea was significantly lower than that of the original flour, indicating significant fractionation of the phenolic compounds into the fine fraction. This agreed with previous studies indicating higher phenolic contents in fine fractions from wheat (Cammerata et al., 2021) and beach pea (Shahidi et al., 2001).

| Accelerated shelf-life testing of pulse flours and protein concentrates
Accelerated shelf-life testing was conducted on the original pulse flours and protein concentrates to better understand the onset of rancidity (Table 8). During the milling process, the cotyledon cellular structure is completely disrupted, increasing the accessibility of lipids to endogenous enzymes associated with rancidity, such as lipase and lipoxygenase (Dundas et al., 1978). At the time of receival, for each pulse type, the protein concentrates were scored higher for rancidity compared with the original flours. The rancidity score for protein concentrates was lowest for faba bean (7%), followed by yellow pea (13%) and red lentil (20%), with a similar trend observed for the original flour samples. Differences in the initial scores between protein concentrates and the original flours may partly be attributed to protein concentrates having a higher fat content, as the fat coconcentrated with protein in the fine fraction during air classification.
The particle size of the protein concentrates is also finer than for the original flours, imparting a different mouthfeel and texture, which can also influence sensory perceptions. Sensory assessment indicated that the onset of rancidity occurred after 3 months of storage (highlighted in bold). The rancidity scores for protein concentrates were found to be highest for faba bean (40%), followed by yellow pea (34%) and red lentil (26%). Sensory evaluation was not continued beyond 3 months of storage, as the panellists found the samples to have a strong and unacceptable aroma at this time. Rancidity is usually accompanied by the formation of FFA, with acidity generally being noticeable to the palate when FFA (calculated as oleic acid) ranges from 0.5% to 1.5%.
In this study, the FFA (0.1%) and pH values (6.3-6.5) remained consistent for all pulse samples, with a small increase in water activity observed after 3 months of storage. However, sensory panels have been found to detect rancidity in extruded chickpea-sorghum snacks before changes in chemical markers (peroxide values) were detected (Bekele et al., 2020).
The resulting data indicate that the perceived onset of rancidity for pulses is more complex than just relying on the formation of FFA, with sensory evaluation encompassing a wide array of aromas and flavours. Pulses contain volatile compounds that are associated with undesirable aromas and off-flavours, which can limit their use in food applications. These off-flavours are often described as green, earthy, bitter and astringent and are partly inherent to the type of pulse but can also develop at various stages across the supply chain, from harvesting through to processing and storage (Roland et al., 2017). Consequently, the limited shelf-life and stability of raw pulse flours and protein concentrates are currently a major constraint for food production and storage. This study demonstrates the importance of developing effective treatments, applicable for dry processing methods, which can extend the shelf-life and stability of protein concentrates for domestic and export markets.

| CONCLUSIONS
The objective of this study was to increase the knowledge available on dry processing of protein concentrates from Australian faba bean, yellow pea and red lentil. Optimal milling and air classification settings were established at pilot scale and will inform domestic commercial scale pulse protein processing operations. However, optimal settings at commercial scale will ultimately depend on the economic significance placed upon protein concentration versus protein yield. The AAS for all three pulse protein concentrates highlighted deficiencies in the sulphur-containing amino acids (MET + CYS) and TRP, with the faba bean protein concentrate also being deficient in THR. These amino acid deficiencies can be addressed at the food production stage, through blending of pulse protein concentrates with complimentary sources of cereal proteins. This could be conducted using selected cereals that would maintain the gluten-free status of pulses, or with cereals that are not gluten-free, but provide functional and T A B L E 8 Accelerated shelf-life testing and sensory assessment of the original pulse flours and protein concentrates Abbreviations: a w , water activity; FFA, free fatty acids. a Sensory assessment was conducted by six panellists, who were instructed to score the degree of perceived rancidity on 15 cm linear scales. The score was converted to a percentage value, in which the higher the score, the more rancid the sample. Scores highlighted in bold were assessed as being rancid.
textural properties for targeted end-use applications, such as plantbased meat alternatives. Plant breeding initiatives could also be tailored towards developing new and improved varieties with higher protein content and more balanced amino acid composition. The coconcentration of protein, ash (minerals), fat, TDF and bioactive constituents, in the fine fractions for all pulses, could lead to the production of protein concentrate ingredients with enhanced nutritional properties, which may potentially impart a range of health benefits.
This may prove to be an additional advantage of dry processing over wet processing methods. Also of significance was the limited shelf-life of protein concentrates, which provides a major constraint for food production and storage. This study demonstrates the importance of developing effective treatments, being suitable for dry processing, that will extend the shelf-life and stability of protein concentrates for domestic and export markets. Furthermore, exploring innovative opportunities to increase the value and end-use applications of the starch concentrate (by-product), beyond the traditional animal feed uses, will lead to further economic gains and increase the overall sustainability of commercial pulse protein processing operations.