HSOA Journal of Dairy Research & Technology Feasibility of Manufacturing a Variety of Soft Cheeses from Milk Separated According to its On-Line Measure of Curd Firmness from Large Sized Commercial Dairy Farms

Feasibility of Manufacturing a of Soft from Milk Separated According to its On-Line Measure of Curd Firmness from Large Sized Com- mercial Dairy Farms. 005. Abstract The feasibility of on-line separation of high-clotting milk or low-clotting milk collected from large sized dairy herds was tested for yield by producing a variety of cheeses. Milk was collected by a large dairy plant and Cottage cheese (5% fat) or soft Israeli cheeses with various percent fats: 5% or 16% were produced. High significant differences ( P<0.001 ) were found for fat, protein and curd firmness at 60 min in all the high-clotting milk and low significant differences ( P<0.001 ) in the low-clotting milk for the same level of milk constituents. Significant differences were found between the different milk separation in the levels of protein, curd firmness at 60 min with or without added Ca, weight of curd and in the calculated protein efficiency. The results of this study indicated an increase in extra protein efficiency (yield) in the different cheeses produced, i.e., 5% Cottage cheese +11%, 5% soft cheese +14% and for 16% soft cheese +7.1%.


Introduction
The modern dairy industry produces a variety of products including drinking milk, fermented products and a large variety of cheeses. Thus, animal breed, nutrition and management do not focus on production of specific or product-designated milk, but rather on maximizing milk volume to maximize revenue in respect to existing payment schemes. The major responsible factors that affect low quality of cheese-milk is associated with intramammary infection with different bacteria species, post intramammary infection, days in milk, stage of lactation, i.e., late or early lactation and estrus [1][2][3]. However, other factors such as non coagulating milk were also associated with low quality milk for cheese making. It was reported that about 8% and 13% of Finnish Ayrshire cows produced non coagulating milk owing to various genetic traits [4,5]. Moreover, studies showed that about 8-9% of milk samples that did not coagulate and that 17-20% of the milk coagulated poorly due to κ-casein and β-lactoglobulin genotypes [6] and that 2% of Danish Holstein and 16% of Swedish Red cows were classified as producing non coagulating milk, which was due to various genotypes [7].
In recent years our team has introduced and presented experimental results on the potential on-line segregation of milk for cheese manufacturing according to its coagulation properties without any addition or alteration of the milk [8,9]. In a later publication, we have discussed the options of making cheese from milk separated in different ratios [10] and recently, we further presented the higher cheese yields obtained from separated milk and discussed the validation of van Slyke's yield equation for calculating cheese yield by the dairy industry [11].The main results have shown that the level of milk constituents changes throughout the milking session along with the milk's coagulation properties. Installing the on-line milk channeling system (Afimilk MCS, Afimilk, Afikim, Israel) enabled separation of milk according to its measured coagulation properties (Afi-CF) and made it possible to divert high-clotting milk into a designated tank and thus to increase cheese yield [10,11]. The Afmilk MCS system is installed in every stall in the milking parlor and segregates every 200 mL pull into high-or low-clotting milk with the cutoff level being decided according to the volume and products to be later produced by the dairy. The system relies on a set of two reception raw milk tanks that are fed by two separated milk lines [8,9]. Despite the need for higher volumes of water and additional labor to handle, clean and haul the milk of the AfiMilk MCS system, economical calculations performed, show a major saving in the total energy consumption owing to higher cheese yield [12].
The current work presents data that show the feasibility of manufacturing a variety of cheeses on a large scale from milk segregated for increased cheese yield. Since the process and yield data are the propriety of the dairy (which was not released to us), only the result of the skimmed milk cheese prepared according to the dairy manufacturing protocol and the protein efficiency for Cottage cheese (5%) or an Israeli variety of soft cheese with 5% or 16% percent fat are presented, in addition to results of cheese made from the same milk on a laboratory scale.

Farms and milk collection
Milk separation was performed in four commercial dairy farms of 350 to 700 Israeli Holstein cows producing on the average >12,500 L milk during 305 days during two years and the different milk streams were transferred to a commercial dairy, which produces ~5,000 tons of Israeli soft cheeses per year. All farms had a double-sided herringbone milking parlors equipped with AfiMilk MCS system with an additional milk line and a releaser, which diverts the milk into two milk tanks A & B. Tank A is designated for high-clotting milk selected for cheese making while the other -Tank B, is intended for producing fluid milk and fermented products. The cows were milked thrice daily at 05:00, 13:00 and 20:00 h, and were fed a typical Israeli total mixed ration containing 65% concentrate and 35% forage (17% protein). Feed was offered ad libitum in mangers located in the sheds. During 2013-2014, 686 batches of Non-Separated Milk (NSM) and 1149 batches of separated milk, i.e., High-Clotting Milk (HCM) or Low-Clotting Milk (LCM) were collected for the experiments. In the first experiment, milk analyses were made on the raw milk, while in the second and third experiments, analyses were made on the pasteurized milk prepared for the various cheeses and the fresh cheeses produced at the laboratory.

Experiment 1
Milk for cheese making was collected rotationally from different farms: NSM, HCM or LCM (Table 1). Milk separation alternated among the 4 farms and each separation cut-off level ratio of HCM to LCM: 20-80, 30-70, 40-60, 50-50 or 60-40 was collected at the same time. Milk from the 4 selected farms was delivered to the dairy by separate tankers. NSM was collected from the same farms on in-between days and the milk was delivered to the dairy as above. Upon arrival at the dairy, milk was sampled from every tanker into 200 mL test cups. Samples were analyzed for gross milk composition: fat, protein and lactose by the Milkoscan FT+ and for Somatic Cell Count (SCC) by the Fossomatic FC (Foss Electric, Hilleröd, Denmark) at the Israel Cattle Breeders Association (ICBA, Caesarea, Israel). A second set of samples comprised 5 L milk of NSM, HCM and LCM in various ratios, of which 'laboratory cheese' was made in designated containers at the Afimilk laboratory (Afikim, Israel). In addition, the milk was tested for clotting parameters: Rennet Clotting Time (RCT) and curd firmness at 60 min (CF60; V) by the Optigraph © (Ysebaert, Frepillon, France). Ten mL samples were placed in the 30 ºC equilibrated wells and Fromase 15 TL (Gist-Brocades nv, Delft, The Netherlands) was added after dilution to achieve clotting at about 600 sec. All samples were analyzed in triplicates.

Experiments 2 & 3
Milk separation alternated among the 4 farms with the decision on cut-off level made by the dairy. Control milk (NSM) was collected from the same farms on in-between days and the milk was delivered to the dairy by separate tankers. For Cottage cheese (5%) milk was sampled from 51 randomly collected batches of NSM milk and 16 (HCM 30), 19 (HCM 40) and 17 (HCM 50) batches of separated milk collected during 2013. Similarly, 51 batches of NSM milk and 68 batches of separated milk (HCM 50) were collected during 2014 from 4 farms. At the dairy, milk was skimmed, pasteurized and samples of ~5 L were taken to Afimilk laboratory for cheese making and to the ICBA laboratory for milk composition tests. The Afimilk laboratory-produced cheese was tested for dry matter and whey was collected and weighed for yield calculation. The skimmed milk was prepared according to the dairy manufacturing protocol for Cottage cheese (5% fat) (Second experiment) 30, 40 50 HCM, or soft Israeli cheeses (Zfatit and Hemed) with various percent fat: 5% or 16% (Third experiment) at 40 HCM.

Cheese making at the laboratory (Laboratory cheese)
Cheese making at the laboratory was performed in triplicates, as described in details [11]. Shortly, six 1-L stainless steel containers were placed in a thermostatically controlled water bath. The milk was preheated for 25 min at 30°C. According to the industrial procedure, a solution of 61.76 g/L CaCl 2 was added after 10 min heating to achieve a final CaCl 2 concentration of 0.00 to 0.2038 g/L (later denoted as +Ca). Maxiren 600 (DSM Food Specialties B.V., Delft, the Netherlands) at 0.089 g/L was added to each container and incubated for 60 min till cutting into 0.8-cm cubes by stainless steel knives. The cut curd was left to stabilize for 10 min and then the temperature was raised to 40°C and cooked for additional 25 min with gentle stirring. The curd was poured into perforated molds and turned over after 10 min. The cheese stored pressed at 46 g/cm 2 for 24 h at 4°C and then weighted for yield calculation. Dry matter in cheese was determined according to standard methods [13].

Statistical Analysis
A univariate model was designed with a logistic model statement using the MIXED procedure of SAS [14], with result outcome as the dependent variable. Three experiments were conducted. The first experiment was analyzed in four steps: a) The independent variable MROUND levels of percent separation (20,30,40,50,60) was analyzed using data from HCM group (n=574) of milk type with the general form: dependent variable (fat, protein, lactose, SCC (×10 3 ), % casein (percent casein from protein), RCT, CF60)=Mround + error.
b) The independent variable Mround levels of percent separation (40,50,60,70&80) was analyzed using data from LCM group (n=574) of milk type with the same model as above.
c) The entire data set (n=686 + 574 +574) was analyzed for differences between milk type groups in all variables using the model described above. Two models were conducted: the 1 st analyzed the differences for all variables using data from HCM and NSM group milk type and the 2 nd analyzed the differences for all variables using data from LCM and NSM group milk type.
B) The entire data set (n=208) was analyzed for differences between HCM and NSM, separately for 2013 and 2014, for all variables described in step A by using the model described in the first experiment.
The third experiment analyzed the pasteurized skimmed milk prepared by the dairy for each type of soft cheese with various percent fat; i.e., 5% or 16%. The dependent variable in pasteurized milk (fat, protein, lactose, SCC (×10 3 ), % casein, RCT, RCT+Ca, CF60, CF60+Ca, curd at 1h and curd at 24h (% moisture, dry matter (g)) and protein efficiency were analyzed for differences between HCM of percent separation(40) (5% fat, n=24; 16% fat, n=32) and NSM (5% fat, n=19; 16% fat, n=28) using the model described in the first experiment. Other parameters in the study such as farm and amount of milk in batch were found with no significant effect and thus removed from the final model. Results are presented as LSmeans + SEM. Protein efficiency was calculated as: (0.83 × % protein in milk)/(1-% dry matter)/100).

Experiment 1
The setup for milk separation was calculated for the 4 farms according to the Afi-CF level determined by the Afilabs in each farm a day before each setting [11]. After deciding on the cutoff level, which results in the ratio of HCM to NSM milk, the actual percent of separated milk was determined. The differences between the calculated separation and the actual ones were ±1-4% (Table 1). The actual NSM was used as a reference and appears as 'Reference' in tables 2 & 3.
Significant differences (P<0.001) were found for fat, protein and CF60 in all the HCM and in the LCM for the same milk constituents. The distribution of protein in the collected milk according to NSM, HCM and LCM is presented (Figure 1). No significant differences among the three milk streams were found for % casein, lactose, SCC and RCT (Table 1). Analysis of those parameters according to the separation level, separately for HCM and LCM milk, revealed no significant differences except for HCM 20 which was significantly higher than HCM 30-60.

Experiment 2
Results of the various parameters measured of skimmed pasteurized milk and laboratory cheese preparation (3 repetitions of 1 L) are summarized ( Table 2). No significant differences were found between NSM of 2013 and 2014 and separated milk HCM (50%) of 2013 and 2014 from the 4 farms. No significant differences were found in all the parameters tested among the three-separated levels of HCM: 30,40,50, therefore, comparison between HCM and NSM was done over HCM levels. Significant differences were found between NSM and HCM milks in the levels of protein, CF60+Ca, weight of curd (wet and dry) and in the calculated protein efficiency in the laboratory test, all of which were significantly higher (P<0.001) for HCM (Table 2). The protein efficiency was ~+4.0 in NSM vs. ~+4.3 in HCM milk. The real protein efficiency of the commercial final product was calculated at the dairy. Accordingly, the extra protein efficiency (by difference) of the final 5% Cottage cheese produced at the dairy was +11% (Table 2).

Experiment 3
Pasteurized skimmed milk prepared by the industrial dairy for soft cheeses (5%) was sampled at the dairy from milk of 24 batches of NSM and 19 (HCM40) batches of separated milk collected from 4 farms during 2014. Similarly, milk prepared for soft cheeses (16%) was sampled at the dairy from 28 batches of NSM milk and 32 (HCM40) batches of separated milk collected from 4 farms during 2014. Average and standard error of pasteurized milk and laboratory cheese (1 L, ×3) are summarized in Table 3. Significant differences were found between NSM and HCM for both cheeses in the skimmed milks in the levels of protein, CF60 and CF60+Ca,weight of curd (wet and dry) and in the calculated protein efficiency in the laboratory test, all of which were significantly higher (P<0.001) for HCM (Table 3).  The protein efficiency for soft cheese (5%) was +4.54 in NSM vs.+4.76 in HCM milk, and for soft cheese (16%) was +4.86 in NSM vs. +5.10 in HCM milk. The extra protein efficiency of the final product at the dairy was +14% of 5% soft cheese and +7.1 of 16% soft cheese (Table 3).

Discussion
Raw milk is produced to fit average production schemes rather than to maximize a specific product, since milk is used for a variety of products with different protocols of industrial manufacturing.     For example, high level of protein in drinking milk is redundant vs. low level of protein in milk for cheese making, which leads to low cheese yield. In modern intensive dairy farms, milk is produced economically when genetics and management maximize milk yield and sold milk volume. Nevertheless, genetic variability among cows, time in lactation, time of milking and milk flow during milking have a great influence on milk constituents. By installing the AfiMilk MCS system that enables separating milk according to its coagulation properties, it was possible to divert part of the milk into a designated tank and thus to increase cheese yield [10,11]. In the first experiment, the AfiMilk MCS system performed at a high efficacy in separating the milk to the pre-decided cutoff level. Unlike earlier results [11] no significant differences in all the parameters tested were found in HCM or LCM milk according to the separation level, except for CF60 in HCM20, which was significantly higher than other HCM levels. The difference was probably related to the number of AfiMilk MCS units, the environment and management differences among farms and time of milk collection during the two years. On the other hand, fat, protein and CF60 in all the HCM target tanks were significantly higher (P<0.001) than in NSM and in all the LCM target tanks these same parameters were significantly lower (P<0.001) than NSM. Other parameters such as % casein, lactose and SCC did not differ significantly from NSM, indicating that the separation is directly related to fat and protein levels but not to the milk quality, which is related to non-and low-clotting milk etc. [4][5][6][7]. Percent casein is the calculation of casein from total protein; therefore, higher protein with similar % casein means higher casein in HCM. Thus, as shown and discussed earlier [11], it is possible to increase dairy cheese production yield by utilizing the same animals and volume of milk for higher profitability. It could be visualized that by using milk channeling systems in large dairy herds for producing HCM and obtaining increased cheese yield per vat volume could result in reducing the number of cows in the herds for producing the same tonnage of cheese.
Many of hard and soft cheeses are prepared from pasteurized milk after skimming the milk to a desired degree of fat to protein ratio. Moreover, in many products, powdered milk or ultrafiltration retentatesare added in order to increase protein content, which results in added costs of concentration and drying [15,16]. Implementation of advanced technological systems, such as the AfiMilk MCS in milking parlors, modifies not only the on-farm management but mostly the procedures practiced by the cheese making plant. Using targeted HCM enables the dairy factory that produces traditional cheeses to skip the standardization of the production formula and only to reduce the fat level in the cheese milk. These modifications broad about ~44% saving of energy usage and ~69% of the carbon dioxide emissions [12].
Due to the limitations of dairy intellectual property, the results of Cottage cheese (5%) (2 nd experiment) or soft cheeses with various percent fat: 5% or 16% (3 rd experiment) are those to the laboratory tests, including laboratory cheese (48 h) produced from milk prepared by the dairy. The extra protein efficiency calculated by the dairy for the final products is also presented. In the second and third experiments, protein, CF60 and the weight of curd were significantly higher in all the HCM milk preparations for the three chesses. As a result, the calculation of the protein efficiency in the laboratory tests (+5%) as well as in the extra protein efficiency of the dairy final product was significantly higher (+7-12%). These results indicate that the potential benefit of the separated milk is manifested in a verity of products with different milk separation levels as was shown by Todde et al. [12].

Conclusion
The AfiMilk MCS system opens new avenues for cheese production plants. Despite the need of installing a second parallel milk line and an additional reception tank, the increased cost of milk hauling and milking parlor cleaning costs diminish when compared to the additional income from the increased cheese yield. As was summarized in regard to the whole scheme: "The implementation of PLF (Precision Livestock Farming) technologies increased energy requirement and carbon dioxide emissions in dairy farms and in the collection of milk, however, the large amount of energy saved in the cheese factory and the increase in cheese production yield make these technologies respectful to the natural resources and to the environment, avoiding about 2.65 MJ of primary energy for every 100 kg of processed milk" [12]. The essence of segregating milk on-line according to its composition and properties opens a new era of transition to dairy farms that produce high value-added product made of the same raw milk, while are also promising in the possibility of reducing the number of animals grown for the same cheese tonnage production.