Moisture Sorption Isotherms Of Feedstuffs And Feed Pellet Cooling Process Modeling

Feed safety as a sector in the food safety system has drawn dramatic public attention; hence moisture control plays an important role in feed quality, feed production margin, animal producers'net profit and the daily consumption of meat, milk and eggs. Stable feed moisture content, esp. in pellet feed, is still a challenge for the global feed industry. Moisture sorption isotherms, which is well known and investigated practically in food industry, is not carrying sufficient awareness it deserves. A relatively systematic research in moisture adsorption and desorption in feed materials and/or compound feeds is absent, thus elementary sorption data and typical predictive equations are deficient. Moisture adsorption and/or desorption equilibration determinations of more than forty feed raw material samples and three typical feed feeds were carried out to find options to improve feed water binding capacity and to reduce water activity, together with choice sorption isotherm equations either original or modified by the authors. A small thin-layer cooling system based upon PAC was established for lab study and online counterflow cooling data acquisition to investigate the possibility to simulate the cooling process by empirical equations and/or artificial neural networks. The research provided preliminary dada and methods to control hot feed pellet cooling practice.1. Adsorption and/or desorption equilibrium moisture content (EMC) of feedstuff samples, including corn, regular soybean meals (SBM), dehulled soybean meal (DHSBM), fermented soybean meal(FSBM), white soybean flakes(WSF) heated at different condition, soybean protein isolates(SPI), cotton seed meal (CSM), degossypololized high-protein CSM(DGCSM), rapeseed meal(RSM), fishmeal(FM), spray-dried blood cells (SDBC), meat and bone meal(MBM), enzymatic hydrolyzed feather meal(HFM), dried beer brewery yeast(DBY), dried distillers'grains and solubles (DDGS), soybean molasses(SBML), sugar beet molasses(BML), choice konjac meal(KJM), soluble starch, gel, and dried ground corn, dried corn cob meal, dried wheat bran and ground rice hull were determined with conventional static gravimetric method with saturated salt solutions and/or the water activity determination instrument, at typical room temperatures (T=25℃, or between 5℃to 45℃). Some of the EMC-Temperature-Aw (water activity, or ERH, equilibrium relative humidity) relationship was modeled. EMC is known as one of the approaches to find the water binding capacity (WBC). This EMC-Aw-Temperature relationship is valuable in feed drying/cooling, handling and storage. In our research, highest WBC were found in quality DDGS, molasses, FSBM and KJM, especially within high water activity range. Water soluble part (WSP) in DDGS linearly increases EMC level, which means new parameters for DDGS selection and quality management. Blahovec and Yanniotis (BY) model with WSP as a variable was the best equation fitting the DDGS EMC at 25℃. Fermentation seemingly increases EMC in soybean meal. Increasing protein solubility of SBM gave a tendency of higher EMC, but did not improve it significantly within the normal solubility range of regular meal, unless in extremely undercooked or overcooked SBM. KJM has very high EMC level, and is inhibitive in mold multiplication. It did not provide higher EMC than the weighted average in mixture with corn. Both the two molasses showed obviously different ERH-EMC relation from other materials, for the high EMC at low ERH range lower than that by saturated MgCl2 solution. Sugar beet molasses took more moisture than soybean molasses. Increasing temperature resulted in higher EMC from ERH provided by saturated BaCl2 solution. Within the animal protein resources, MBM and HFM had lower EMC than fish meal and SDBL. The fitness of Aw-EMC relationship with selected equations could be helpful in fish meal adulteration detection. High protein CSM showed lower EMC than low protein meal. When Aw is under 0.843, RSM EMC is lower than all the four CSMs, but from 0.843 it is very close to that of CSMs. This partially explains the lower EMC in grownout grass carp feed than the level in the broiler feed and the duck feed. Particle size did not affect the EMC of corn and SBM, though the finest corn had seemingly lower EMC, possibly due to the low initial moisture in this fine corn. Ground rice hull is a good premix carrier or dilutor for premix, judged by the Aw-EMC relation and the moisture-flowability interaction. New sorption isotherm equations including Peleg, Generalized D'Arcy and Watt (GDW), Blahovec and Yanniotis (BY) as well as the conventional GAB model fit generally all the materials except for molasses, while other conventional models such as Henderson, Oswin, Halsey and Chung-Pfost performed well for specific feedstuffs and/or specific water activity ranges, e.g. mChung-Pfost for corn desorption process and mHalsey for cottonseed meals. Peleg and mPeleg described well the ERH-EMC-T relationship of molasses when ERH<=84% from 15℃to 45℃.2. Moisture sorption behavior of pelleted finishing broiler feed, pelleted finishing duck feed and pelleted grownout grass carp feed differed from each other, within the temperature range from 15℃to 45℃. Duck feed and grass carp feed absorbed much higher moisture at high temperature and humid condition (45℃, ERH>90%), which did not follow the principle of normal agricultural products that increasing temperature reduces the EMC. The model Peleg, GDW, BY, GAB and GAB-VR were choice models for the feed pellets. Henderson, Oswin, Halsey and Chung-Pfost as well as their modified forms and Chen-Clayton model either fits secondary to the choice or have limited generalization and utilization. Also as stated in feed material sorption research, this compound feed sorption study successfully introduced temperature effect as a new variable into either Peleg, GDW or BY model, which was approved by published data. These models were adopted to estimate proper moisture content for safe storage, with Aw=0.65 as the critical Aw point. Traditional sorption isotherm equations were more convenient for latent heat calculation once they fit the data with accepted goodness.3. A thin layer cooling data acquisition system was established to facilitate the approval of possibility to find empirical parametric equations and/or artificial neural network (ANN) models that integrated all key independent variables for predicting and controlling the cooling process. Thin layer cooling models lack only pellet flow, temperature gradient and relative humidity gradient that exist in deep bed cooling. This system and data provided information for layer by layer cooling simulation after dividing the pellet bed into thin-layers with proper depth increment. Empirical equations with 15 to 22 parameters were obtained for fitting the time to a pellet temperature ten (T10) and five degrees Celsius (T5) above air and for fitting the moisture content at these two time points (M10, M5), of the pelleted duck feed during thin-layer cooling at various pellet initial condition (temperature from 48℃to 87℃,moisture from 12.67% to 23.54% db) and ambient condition (air temperature from 7℃to 40℃, relative humidity from 19% to 94%, and air velocity from 0.28 m/s to 2.88 m/s). EMC calculated by choice sorption equation(s) was involved in the regression. The empirical moisture equation with cooling time as an independent variable performed better than the time equation. The cooling time equation including average moisture loss rate gave much better fitting goodness than the equation without the rate. If the moisture loss rate was excluded the cooling time model gave a fitting accuracy close to acceptance, but the predictive accuracy for the validation data group required improvement. Equalization of variables did not improve the fit significantly. T5 could be linearly estimated by T10 (T5≈1.52T10) and M5 linearly by M10 (M5≈0.99M10). For moisture content, ANN gave much better prediction and validation even without the cooling time variable, than the established empirical moisture equation, but neither of them could fit or simulate the cooling time excellently. A combination of the merits of ANN and the empirical parametric model might be another approach. Simulation of the duck feed (assigned initial temperature 80℃and moisture:18% db) at 55% RH from 10℃to 40℃with the empirical equation found that faster air movement reduced exponentially the cooling time. Relative humidity changed the reductive effect. Higher RH increased the moisture content in cooled feed linearly (slope was 0.0371 at 10℃and 0.0247 at 20℃), raised air velocity reduced moisture level and the Hoerl model (M=a*bVair*VairC) described well this function. However, the moisture increasing effect was different at hot environment (30℃～38℃) from that at normal range (10℃～20℃) because on the contrary the velocity higher than 1.444 m/s increased the moisture percentage in the cooled pellets. High air temperature weakened the effect of relative humidity on moisture in the product, thus the cooling configuration requires adjustment from that in lower temperature condition. The temperature and water data obtained by the thin-layer cooling system are useful for calculation of coupled heat and mass transfer coefficients.4. Infrared thermo-videoing system were managed to record the temperature variation in the steam conditioned mash flow at the steam conditioner exit and in the cooled pellets at the cooler exit plain. The average lowest temperature was about 12℃lower than the average highest in the mash flow. Occasional cool and wet clots were found as well. Deviation of temperature and moisture of feed at the cooler downloading plain varied from cooler to cooler. These findings will help avoid fluctuation in the planned cooling automation system. The trial suggested extensive potential utilization of thermo videoing in both research and production in feed, food, cereal and oilseed industries.5. A proper ANN network simulated well the online counterflow cooling of a pelleted aqua feed for fresh water farming. Evaluation of variable importance by a multi-layer preceptor network (MLP) provided references for automation design. For the specific product manufactured by the specific processing line within the air circumstances, mash moisture, pellet moisture at the die exit, EMC, conditioned mash temperature and moisture as well as the air temperature were profound factors on product moisture, while the air flow (determined by air valve openings), EMC, conditioned mash moisture, air temperature and the conditioned mash temperature were critical factors on cooled pellet temperature.6. The data obtained on pellet feed EMC and corresponding choice equation(s), the latent heat, and the thin-layer cooling information as well as the moisture effect on pellet diameter, pellet density and bulk density etc. facilitate substantially the numerical solution of mass and heat problems with differentiation equations, problems by ANSYS and/or that in CFD modeling. This study tried to establish and solve numerically the heat and mass balance equation groups with assistance by the collected data. However the report was not good as expected, which means further effort is necessary to improve the modeling, the solving and/or the data base.