Cheese Moisture Research Articles

Novel Online Sensor Technology for Continuous Monitoring of Milk Coagulation and Whey Separation in Cheesemaking

The cheese industry has continually sought a robust method to monitor milk coagulation. Measurement of whey separation is also critical to control cheese moisture content, which affects quality. The objective of this study was to demonstrate that an online optical sensor detecting light backscatter in a vat could be applied to monitor both coagulation and syneresis during cheesemaking. A prototype sensor having a large field of view (LFV) relative to curd particle size was constructed. Temperature, cutting time, and calcium chloride addition were varied to evaluate the response of the sensor over a wide range of coagulation and syneresis rates. The LFV sensor response was related to casein micelle aggregation and curd firming during coagulation and to changes in curd moisture and whey fat contents during syneresis. The LFV sensor has potential as an online, continuous sensor technology for monitoring both coagulation and syneresis during cheesemaking.

Ultrafiltered Milk Reduces Bitterness in Reduced-Fat Cheddar Cheese Made with an Exopolysaccharide-Producing Culture

The objectives were to reduce bitterness in reduced-fat Cheddar cheese made with an exopolysaccharide (EPS)-producing culture and study relationships among ultra-filtration (UF), residual chymosin activity (RCA), and cheese bitterness. In previous studies, EPS-producing cultures improved the textural, melting, and viscoelastic properties of reduced-fat Cheddar cheese. However, the EPS-positive cheese developed bitterness after 2 to 3 mo of ripening due to increased RCA. We hypothesized that the reduced amount of chymosin needed to coagulate UF milk might result in reduced RCA and bitterness in cheese. Reduced-fat Cheddar cheeses were manufactured with EPS-producing and nonproducing cultures using skim milk or UF milk (1.2×) adjusted to a casein:fat ratio of 1.35. The EPS-producing culture increased moisture and RCA in reduced-fat Cheddar cheese. Lower RCA was found in cheese made from UF milk compared with that in cheese made from control milk. Ultrafiltration at a low concentration rate (1.2×) produced EPS-positive, reduced-fat cheese with similar RCA to that in the EPS-negative cheese. Slower proteolysis was observed in UF cheeses compared with non-UF cheeses. Panelists reported that UF EPS-positive cheese was less bitter than EPS-positive cheese made from control milk. This study showed that UF at a low concentration factor (1.2×) could successfully reduce bitterness in cheese containing a high moisture level. Because this technology reduced the RCA level (per g of protein) to a level similar to that in the control cheeses, the contribution of chymosin to cheese proteolysis would be similar in both cheeses.

Effect of Type of Concentrated Sweet Cream Buttermilk on the Manufacture, Yield, and Functionality of Pizza Cheese

Sweet cream buttermilk (SCB) is a rich source of phospholipids (PL). Most SCB is sold in a concentrated form. This study was conducted to determine if different concentration processes could affect the behavior of SCB as an ingredient in cheese. Sweet cream buttermilk was concentrated by 3 methods: cold (<7°C) UF, cold reverse osmosis (RO), and evaporation (EVAP). A washed, stirred-curd pizza cheese was manufactured using the 3 different types of concentrated SCB as an ingredient in standardized milk. Cheesemilks of casein:fat ratio of 1.0 and final casein content ∼2.7% were obtained by blending ultrafiltered (UF)-SCB retentate (19.9% solids), RO-SCB retentate (21.9% solids), or EVAP-SCB retentate (36.6% solids) with partially skimmed milk (11.2% solids) and cream (34.6% fat). Control milk (11.0% solids) was standardized by blending partially skimmed milk with cream. Cheese functionality was assessed using dynamic low-amplitude oscillatory rheology, UW Meltprofiler (degree of flow after heating to 60°C), and performance of cheese on pizza. Initial trials with SCB-fortified cheeses resulted in ∼4 to 5% higher moisture (51 to 52%) than control cheese (∼47%). In subsequent trials, procedures were altered to obtain similar moisture content in all cheeses. Fat recoveries were significantly lower in RO- and EVAP-SCB cheeses than in control or UF-SCB cheeses. Nitrogen recoveries were not significantly different but tended to be slightly lower in control cheeses than the various SCB cheeses. Total PL recovered in SCB cheeses (∼32 to 36%) were lower than control (∼41%), even though SCB is high in PL. From the rheology test, the loss tangent curves at temperatures>40°C increased as cheese aged up to a month and were significantly lower in SCB cheeses than the control, indicating lower meltability. Degree of flow in all the cheeses was similar regardless of the treatment used, and as cheese ripened, it increased for all cheeses. Trichloroacetic acid-soluble N levels were similar in the control and SCB-fortified cheese. On baked pizza, cheese made from milk fortified with UF-SCB tended to have the lowest amount of free oil, but flavor attributes of all cheeses were similar. Addition of concentrated SCB to standardize cheesemilk for pizza cheese did not adversely affect functional properties of cheese but increased cheese moisture without changes in manufacturing procedure.

Effect of resistant starch on the water binding properties of imitation cheese

The effect of moisture and resistant starch concentration on the water activity ( a w) and water sorption of imitation cheese was investigated. The Guggenheim–Anderson–deBoer (GAB) equation was found to be a good description of the experimental moisture sorption isotherms. Inclusion of starch in cheese increased moisture sorption over the a w range studied. The GAB monolayer value indicated starch-containing cheese bound an additional 1.4 g water/100 g dry matter when compared to a starch-free cheese. Increasing cheese moisture content had an insignificant effect on moisture sorption but increased a w. Increasing starch concentration had little influence on the GAB parameters but nonetheless decreased a w. It is believed that starch, while relatively inert in terms of water binding, may organise water molecules causing them to be physically entrapped within the cheese matrix. By not competing with protein for water molecules, the starch allowed the water to remain available for plasticization to maintain good functionality of the cheese.

Replacement of fat with functional fibre in imitation cheese

The physical properties of imitation cheeses containing 21.3% dry matter (DM) resistant starch with moisture levels of 52–60%, as well as cheeses (60% moisture) containing 21.3–43.2% DM resistant starch were examined. Flowability of the heated cheese was assessed by an empirical method; the textural properties by texture profile analysis and dynamic rheology. Sensory analysis of cheeses was conducted by panellists using ranking tests. Increasing the moisture decreased cheese hardness did not affect flow but led to an undesirable ‘sticky’ and ‘clingy’ mouthfeel. Hardness increased linearly with increasing starch. Flow decreased at starch contents ⩾34.8% DM. Replacement of fat in the higher moisture cheeses resulted in the production of a good, firm cheese as assessed by a sensory panel. It was possible to incorporate up to 43.2% DM resistant starch into cheeses, a level higher than many ‘high fibre’ products, while maintaining acceptable functional properties.

Effect of Protein-to-Fat Ratio of Milk on the Composition, Manufacturing Efficiency, and Yield of Cheddar Cheese

Twenty-three Cheddar cheeses were prepared from milks with a protein content of 3.66% (wt/wt) and with different protein-to-fat ratio (PFR) in the range 0.70 to 1.15; the PFR of each milk differed by 0.02. For statistical analysis, the 23 cheeses were divided into 3 PFR groups: low (LPFR; 0.70 to 0.85), medium (MPFR; 0.88 to 1.00) and high (HPFR; 1.01 to 1.15), which were compared using ANOVA. The numbers of PFR values in the LPFR, MPFR, and HPFR groups were 9, 7, and 7, respectively. Data were also analyzed by linear regression analysis to establish potentially significant relationships among the PFR and response variables. Increasing PFR significantly increased the levels of cheese moisture, protein, Ca, and P, but significantly reduced the levels of moisture in nonfat substances, fat-in-DM, and salt-in-moisture. The percentage of milk fat recovered in the LPFR cheese was significantly lower than that in the MPFR or HPFR cheeses. In contrast, the recovery of water from milk to the LPFR cheese was significantly higher than that in the MPFR or HPFR cheeses. Increasing the PFR led to a significant decrease in the actual yield of cheese per 100kg of milk but a significant increase occurred in the normalized yield of cheese per 100kg of milk with reference values of fat plus protein (3.4 and 3.3%, wt/wt, respectively). The results demonstrate that alteration of the PFR of cheese milk in the range 0.70 to 1.15 has marked effects on cheese composition, component recoveries, and cheese yield.

Interaction of Brine Concentration, Brine Temperature, and Presalting on Salt Penetration in Ragusano Cheese

Thirty-one 3.6-kg blocks of Ragusano cheese were made on each of 6 different days (in different weeks) starting with a different batch of milk on each day. On d 1, 3, and 5, the cheeses were not presalted and on d 2, 4, and 6, all cheeses were presalted (PS). One of the 31 blocks of cheese was selected at random for analysis before brine salting (i.e., on d 0). The remaining 30 blocks were randomly divided into 2 groups of 15 blocks each; one group was placed in 18% brine (18%B) and the other group was placed in saturated brine (SB). For the 15 blocks within each of the 2 brine concentrations (BC), 5 blocks were placed in a brine tank at 12° C, 5 at 15° C, and 5 at 18° C, and submerged for 24 d. The research objective was to determine the combined impacts (i.e., interactions) of PS the curd before stretching, BC (SB vs. 18%B), and brine temperature (BT; 12, 15, and 18° C) on salt uptake, moisture content, and yield of Ragusano cheese. Although BC, BT, and PS each had their own separate impacts on salt uptake, there was little interaction of these effects on salt uptake when they were used in combination. The PS most quickly delivered salt to the interior of the cheese and was the most effective approach to salting for controlling early gas formation. There were strong separate impacts of BC, BT, and PS on cheese moisture content, moisture loss, and net weight loss, with BC having the largest separate impact on these parameters. Reducing BT reduced salt content and increased moisture, but the effects were small. The more important effect of reduced BT was to reduce growth of gas forming bacteria. The 18%B produced higher moisture, and less moisture and weight loss than SB. The effect of interactions of BC, BT, and PS on moisture loss and net weight loss were small. To achieve the maximum benefit from the various approaches to salting for controlling early gas formation in Ragusano cheese, PS combined with slightly lower BT (i.e., 15° C instead of 18° C) should be used. Although using 18%B instead of SB did increase salt uptake, the point at which improved salt uptake occurred due to use of 18%B did not provide benefit in prevention of early gas formation, as reported separately. However, use of 18%B instead of SB provided a 9.98% increase in cheese yield due to reduced moisture loss during brining; this would be very attractive to cheese makers. The increase in yield needs to be balanced against the risk of growth of undesirable bacteria in the 18%B and the creation of another cheese quality defect.

Influence of Condensed Sweet Cream Buttermilk on the Manufacture, Yield, and Functionality of Pizza Cheese

Compositional changes in raw and pasteurized cream and unconcentrated sweet cream buttermilk (SCB) obtained from a local dairy were investigated over 1 yr. Total phospholipid (PL) composition in SCB ranged from 0.113 to 0.153%. Whey protein denaturation in pasteurized cream over 1 yr ranged from 18 to 59%. Pizza cheese was manufactured from milk standardized with condensed SCB (∼34.0% total solids, 9.0% casein, 17.8% lactose). Effects of using condensed SCB on composition, yield, PL recovery, and functional properties of pizza cheese were investigated. Cheesemilks were prepared by adding 0, 2, 4, and 6% (wt/wt) condensed SCB to part-skim milk, and cream was added to obtain cheesemilks with ∼11.2 to 12.7% total solids and casein:fat ratio of ∼1. Use of condensed SCB resulted in a significant increase in cheese moisture. Cheese-making procedures were modified to obtain similar cheese moisture contents. Fat and nitrogen recoveries in SCB cheeses were slightly lower and higher, respectively, than in control cheeses. Phospholipid recovery in cheeses was below 40%. Values of pH and 12% trichloro-acetic acid-soluble nitrogen were similar among all treatments. Cheeses made from milk standardized with SCB showed less melt and stretch than control cheese, especially at the 4 and 6% SCB levels. Addition of SCB significantly lowered free oil at wk 1 but there were no significant differences at wk 2 and 4. Use of SCB did not result in oxidized flavor in unmelted cheeses. At low levels (e.g., 2% SCB), addition of condensed SCB improved cheese yield without affecting compositional, rheological, and sensory properties of cheese.

Effect of Milk Protein Standardization Using Different Methods on the Composition and Yields of Cheddar Cheese

Two sets of Cheddar cheese were made in which the milk protein level (%, wt/wt) was increased from 3.3 (Control A, CA) to 3.6 (set A) or from 3.3 (control B, CB) to 4.0 (set B) by the addition of phosphocasein (PC), milk protein concentrate (MPC), or freshly prepared ultrafiltered milk retentate (UFR). The cheeses were denoted CA, PCA, MPCA, and UFRA from set A, and CB, PCB, MPCB, and UFRB, from set B, respectively. The level of cheese moisture decreased significantly on increasing milk protein level from 3.3 to 3.6 or 4.0% (wt/wt), but was not affected significantly by the method of protein standardization. The percentage milk fat recovered to cheese increased significantly on increasing the level of milk protein from 3.3 to 3.6% (wt/wt) with PC, and from 3.3 to 4.0% (wt/wt) with PC, MPC, and UFR. Increasing milk protein level from 3.3 to 4.0% (wt/wt) with PC significantly increased the percentage of milk protein recovered to cheese. Actual cheese yield increased significantly with milk protein level. The yield of cheese per 100kg of milk normalized to reference levels of fat (3.4%, wt/wt) and casein (2.53%, wt/wt) indicated no significant effects of protein content or standardization treatment on yield. However, the moisture-adjusted yield per 100kg of milk with reference levels of fat and casein increased significantly on increasing the protein content from 3.3 to 3.6% (wt/wt) with MPC and from 3.3 to 4.0% (wt/wt) with PC, MPC, and UFR.

The study of the grossing process in the grosses of fluted type

This paper gives information about device of gross for fabrication of polymeric fluted sheet, capacity of grosses, the experimental study of the hydrodynamic characteristic of the melt form of the polymeric materials and the characteristic features of toughness of the fluted sheet. Authors gave special emphasis on reological features, factor of volume increasing, moisture cheese, toughness at spraining and density fluted sheet. © 2006 Wiley Periodicals, Inc. J Appl Polym Sci 103: 72–80, 2007

Improvement of Texture and Structure of Reduced-Fat Cheddar Cheese by Exopolysaccharide-Producing Lactococci

The objective of this study was to evaluate the effect of capsular and ropy exopolysaccharide (EPS)-producing strains of Lactococcus lactis ssp. cremoris on textural and microstructural attributes during ripening of 50%-reduced-fat Cheddar cheese. Cheeses were manufactured with added capsule- or ropy-forming strains individually or in combination. For comparison, reduced-fat cheese with or without lecithin added at 0.2% (wt/vol) to cheese milk and full-fat cheeses were made using EPS-nonproducing starter, and all cheeses were ripened at 7°C for 6 mo. Exopolysaccharide-producing strains increased cheese moisture retention by 3.6 to 4.8% and cheese yield by 0.28 to 1.19kg/100kg compared with control cheese, whereas lecithin-containing cheese retained 1.4% higher moisture and had 0.37kg/100kg higher yield over the control cheese. Texture profile analyses for 0-d-old cheeses revealed that cheeses with EPS-producing strains had less firm, springy, and cohesive texture but were more brittle than control cheeses. However, these effects became less pronounced after 6mo of ripening. Using transmission electron microscopy, fresh and aged cheeses with added EPS-producing strains showed a less compact protein matrix through which larger whey pockets were dispersed compared with control cheese. The numerical analysis of transmission electron microscopy images showed that the area in the cheese matrix occupied by protein was smaller in cheeses with added EPS-producing strains than in control cheese. On the other hand, lecithin had little impact on both cheese texture and microstructure; after 6mo, cheese containing lecithin showed a texture profile very close to that of control reduced-fat cheese. The protein-occupied area in the cheese matrix did not appear to be significantly affected by lecithin addition. Exopolysaccharide-producing strains could contribute to the modification of cheese texture and microstructure and thus modify the functional properties of reduced-fat Cheddar cheese.

Influence of Calcium, pH, and Moisture on Protein Matrix Structure and Functionality in Direct-Acidified Nonfat Mozzarella Cheese

Influence of calcium, moisture, and pH on structure and functionality of direct-acid, nonfat Mozzarella cheese was studied. Acetic acid and citric acid were used to acidify milk to pH 5.8 and 5.3 with the aim of producing cheeses with 70 and 66% moisture, and 0.6 and 0.3% calcium levels. Cheeses containing 0.3% calcium were softer and more adhesive than cheeses containing 0.6% calcium, and flowed further when heated. Cheeses with the same calcium content (0.6%), the same moisture content, but set at different pH values (pH 5.3 and 5.8), exhibited no significant differences in melting or firmness. Increasing cheese moisture content from 66 to 70% produced a softer cheese but did not increase meltability. Such differences in functionality corresponded with differences in structure and arrangement of proteins in the cheese protein matrix. Microstructure of cheese with 0.6% calcium had an increase in protein folds and serum pockets compared with the 0.3% calcium cheeses that had a more homogeneous structure. Protein matrix in the low-calcium cheese appeared less dense indicating the proteins were more hydrated. In the 0.6% calcium cheeses, the proteins appeared more aggregated and had larger spaces between protein aggregates. Thus, between pH 5.3 and 5.8, calcium controls cheese functionality, and pH has only an indirect affect related to its influence on the calcium in cheese.

QUALITY AND SHELF LIFE ENHANCEMENT OF DOMIATI CHEESE FROM HEAT TREATED MILK USING STARTER CULTURE

10%) with adding 1% activated commercial starter cultures of YC-X11 (Str. thermophilus and Lb. delbruekii ss. bulgaricus, T1), Bio Profit (Lb rhamnosus and Propio. freudenreichii ss. shermanii, T2) and LC 705 (Lb. casei, T3). Resulting cheeses were pickled into its own whey. The low-salted cheese was pickled for 6 months and the high-salted cheese after 9 months. Moisture, salt and yield of low and high-salt cheeses decreased with adding starter culture, while an increase was noticed in the acidity, soluble nitrogen (SN) and total volatile fatty acids (TVFA). High salt cheeses had significantly higher values for moisture, salt and yield with significantly lower acidity, SN and TVFA than low salt cheeses. The values of moisture, salt and yield gradually decreased during pickling while acidity, SN and TVFA significantly increased. Domiati cheese with starter culture had higher total bacterial count (TBC) than that of control being highest in T3. Increasing the salt in cheese milk resulted in lower TBC. Values of TBC increased in all samples during early pickling and then sharply decreased till the end. There was a remarkable inhibition in the growth of sporeforming bacteria and yeasts & moulds with adding starter culture. Sensory quality attributes of cheese improved with adding starter culture. Pickling of cheese up to different periods led to better flavour as well as body & texture but extending the period beyond caused lower quality. The rate of improvement was faster in cheese of low salt than that from 10% salted milk. Starter culture of Bio Profit (T2) produced cheese with typical ripened cheese flavour and texture and can be recommended for low or high-salt cheeses.

Prediction of Inorganic Salt and Moisture Content of Process Cheese Using Dielectric Spectroscopy

The development of on-line sensors for compositional analysis during cheese manufacture is desirable for improved quality control. Dielectric properties of a food product are principally determined by its moisture and salt content. This indicates that dielectric spectroscopy may offer a rapid, on-line and non-destructive method for the determination of moisture and salt content of process cheese. However limited information is available in the literature on the dielectric properties of process cheese. Therefore the aims of this study are to investigate the dielectric properties of process cheese samples over a range of compositional parameters and to assess the potential of dielectric spectroscopy to improve process control during process cheese manufacture. Dielectric spectra of process cheese samples were measured using a coaxial line probe between 300 MHz and 3 GHz. A clear tend was observed between higher moisture content and increases in the dielectric constant. Inorganic salt content was found to have a major influence on the loss factor. The dielectric data obtained was used to develop chemometric models for the prediction of moisture and inorganic salt content of two experimental sets of process cheese samples (exp A and exp B). The root mean square error of prediction (RMSEP) for the models developed to predict moisture content were 0.524% (w/w) (exp A), and 0.423% (w/w) (exp B), while the RMSEP of the inorganic salt models were 0.220% (w/w) (exp A), and 0.263% (w/w) (exp B). It was concluded that dielectric spectroscopy has potential application for compositional analysis in process cheese manufacture.

Prediction of moisture in cheese of commercial production using neural networks

Abstract Control of cheese moisture is paramount to maximizing yield and profitability of a cheesemaking operation. Modeling and prediction of cheese moisture prior to pressing from a large industrial database for stirred-curd Cheddar cheese made with non-standardized and standardized milk was carried out using neural networks (NN). The number of model input variables was reduced by removing or combining some of them, based on cheesemaking knowledge and on the results of two tests estimating the impact of each model input. Input removal was carried out until the validation mean absolute prediction error (MAPE) increased. An initial NN cheese moisture model with 38 input process variables, coded as 57 NN inputs, was reduced to one with 21 input process variables, coded as 34 NN inputs. For the latter, the validation MAPE was 0.53% cheese moisture in a range of cheese moisture of 13.2%, and 0.51% for the best 25% of models (out of 100). For the range of operating conditions of the process in this study, four main groups of variables were found to be the most influential on the prediction of cheese moisture: cutting and subsequent stirring of the curd, curd rinsing temperature, starter quantity, activity and strain, and seasonal variation of milk composition. The NN model with the selected input variables and optimized number of hidden neurons was then used to predict cheese moisture for ranges of these variables. This study showed that NN models can successfully extract input–output variable relationships from industrial production data in spite of the inherent error in these data. The resulting NN models can be used both for research to develop the base of knowledge on production variables and their complex interactions, as well as for the prediction of cheese moisture.

Contribution of proteolytic activity associated with somatic cells in milk to cheese ripening

Abstract The effect of proteolytic enzymes from somatic cells on cheese quality was studied. In preliminary experiments, milk and two sodium caseinate systems (pH 6.5 and pH 5.2, the latter in the presence of 5% NaCl) were used as substrates to investigate the proteolytic activity of somatic cells recovered from mastitic milk. Urea-polyacrylamide gel electrophoretograms of hydrolysates suggested that somatic cell extracts contributed directly to proteolysis both in buffer and in milk, but that such activity was reduced by batch pasteurisation (63 °C for 30 min). Sodium caseinate was readily hydrolysed by somatic cell extracts; hydrolysis of α s1 -casein was greater at pH 5.2 and increased with level of somatic cells, suggesting that somatic cells contain proteolytic enzymes which are more active at acidic pH values. Subsequently, miniature Cheddar-type cheeses were made from batches of milk to which somatic cells were added (at levels of levels of 3×10 5 or 6×10 5 cells mL −1 ), either before or after pasteurisation. Proteolysis during ripening of cheese (as measured by levels of pH 4.6-soluble nitrogen) increased with somatic cell addition, although this effect was reduced by pasteurisation after cell addition. Somatic cells may also have directly influenced cheese moisture content, which has been established as a principal indicator of quality of Cheddar-type cheese. Proteolytic enzymes of somatic cells from milk were shown to contribute directly to proteolysis in milk and cheese.

Use of low concentration factor ultrafiltration retentates in reduced‐fat Minas Frescal cheese manufacture: effect on yield and sensory properties

The yield and sensory properties of reduced‐fat Minas Frescal cheese made from low concentration factor (CF) retentates were studied. Three different CFs were tested (1.2, 1.5 and 1.8). The chemical compositions of the milk, retentate, whey and cheese were determined, as well as the cheese yield. The cheese moisture content decreased with increasing CF. The cheese yield was significantly dependent on the CF in the same direction as the moisture content. Despite compositional differences among the samples, only the cheese made with a CF of 1.8 presented low sensorial acceptance. CF 1.2 was found to be the optimum value for reduced‐fat Minas Frescal cheese manufacture in the CF range studied.

A systems analysis of pasta filata process during Mozzarella cheese making

Pasta filata process during Mozzarella cheese manufacturing was studied. System variables (SME, specific mechanical energy; TE, exit temperature) were defined and their relationships to operating conditions were investigated. The compositional, rheological and functional properties of cheeses manufactured under different conditions were analyzed and relationships between system variables and cheese properties were established. The cheese moisture content decreased as SME increased. A quadratic relation was established between cheese moisture content and TE, FDM (fat-in-dry-matter) and SME and FDM and TE. These relationships suggested that optimization of the cheese yield is possible by adjusting the pasta filata process operating parameters. Evaluation of microstructural features of the cheese samples further supported these conclusions.

Comparison of Biogenic Amine Profile in Cheeses Manufactured from Fresh and Stored (4°C, 48 Hours) Raw Goat's Milk

In this study, the evolution of microbial counts, biogenic amine contents, and related parameters (pH, moisture, and proteolysis) in goat cheese made from fresh raw milk or raw milk stored for 48 h at 4°C was examined. In both cases the milk was nonpasteurized. This study was designed to evaluate the effect of milk quality on the profile of biogenic amines in relation to the evolution of the microbial population during cheese making. Cheese made from raw milk stored for 48 h at 4°C showed the highest microbial counts and biogenic amine levels. The storage of milk under refrigeration caused significant increases in the levels of some microbial and biogenic amines during ripening, but not initially. Tyramine was the main biogenic amine in the two cheeses tested, followed by cadaverine. However, the main differences in amine contents between batches were found for putrescine, histamine, and β-phenylethylamine,whose levels were more than twofold higher in samples from raw milk refrigerated for 48 h than in samples from fresh milk.

Impact of modeling parameters on the prediction of cheese moisture using neural networks

In the cheese manufacturing process, yield is of great economic importance, and cheese moisture is the single most important factor influencing yield. A large industrial database of 41 process variables was used to develop a comprehensive model for cheese moisture using three-layer feedforward neural networks (NNs). The scope of this research was to analyze NN modeling parameters related to the process and data, in view of obtaining the best predictive model for cheese moisture. An analysis on the number of hidden neurons and the fraction of data used for on-line validation (FV) showed that the latter influences the choice of the number of hidden neurons and determines how the selected model is representative. For the cheese moisture model, six hidden neurons and 20% FV were optimal. Numerous trials were performed with the objective of reducing the number of NN inputs while aiming for better or equivalent model accuracy. These included the development of specialized models associated to specific values of some qualitative variables, various encodings for qualitative variables, removal of variables to increase the number of modeling data, and combinations of some of these strategies. It was found that NN could be used to derive predictive models for cheese moisture from industrial data, capable of attaining a validation mean absolute prediction error of 0.56% cheese moisture, for data with a range of 13.2% cheese moisture. Trials indicated that gains in accuracy were achieved by using several specialized NN models for each value of a qualitative variable instead of an overall model, as well as by encoding qualitative inputs as quantitative inputs directly related to the output. The validation prediction error exhibited a high correlation with the number of training data per NN coefficient, but was highly influenced by the actual choice of model inputs.