Sensitivity analysis of environmental impact of milk powder spray drying system based on intelligent algorithm

Abstract

The increasing market demand for milk powder has not only promoted the production capacity of milk powder, but also increased the impact on the environment. Therefore, it is very important to study the relationship between the environmental impact of milk powder spray drying (MPSD) system and system-related parameters and identify the key parameters to improve the efficiency of the sustainable improvement of the system. Treed Gaussian Process (TGP) and Standardized Regression Coefficients (SRC)methods are used to analyze the sensitivity of the system to environmental impacts. The results show that the inlet air temperature of the drying tower has the greatest impact on the environment of the system, accounting for about 82%, followed by the atomization pressure and the feed pump speed, accounting for about 9% and 8% respectively. Moreover, not only the environmental performance of the system should be improved, but also the quality of milk powder should be guaranteed when optimizing the parameters such as the inlet air temperature of drying tower. This study can help the manufacturers of milk powder and related equipment to determine the priority of improving the system from the perspective of environmental protection.

Keywords

Sensitivity analysis treed Gaussian process standardized regression coefficients spray drying environmental impact

1 Introduction

Spray drying has been used in the food industry for many years. Due to the low temperature and humidity of the surface of the droplets during the drying process, the color, flavor and cellulose contained in the food can continue to be preserved. Therefore, this process is particularly suitable for processing heat-sensitive foods and has become the main method used by dairy factories to produce various milk powder products [1]. In recent years, the market demand for milk powder has increased year by year, which promotes the production of milk powder, so that the resources consumed in the production process and the impact on the environment are increased with each passing year [2]. Studies have shown that one of the main challenges faced by the food processing industry in terms of process optimization and innovation is to reduce the impact of food production on the environment [3, 4]. Spray drying has the largest energy consumption in the processing of milk powder [4]. Therefore, it is necessary to study the relationship between the environmental impact of milk spray drying (MPSD) system and the relevant parameters of the system and find the key parameters to reduce energy consumption and emissions.

Although there is very little research on the sensitivity of the environmental impact of the MPSD system, some researchers have carried out research on the environmental impact of other driers. Gregory et al. [5] evaluated and compared the environmental impact of five hand drying systems, including hands-under dryers and high-speed hands-under dryers, high-speed hands-in dryers, cotton roll towels and paper towels, and the influence of the uncertainty of input data on the relative performance of the product is also evaluated. The study found that high-speed dryers have less environmental impact than paper towels and cotton roll towels. Chavan et al. [6] made some practical improvements to the solar conductive dryer (SCD) and performed computational fluid dynamics simulations on the improved dryer. It was found that the improved SCD was superior to the existing dryer in terms of outlet wind speed, outlet air temperature, intermediate plate temperature and material handling capacity. Liu et al. [7] studied the effects of drying temperature, vacuum pressure duration (VPD) and ambient pressure duration (APD) on the pulse vacuum drying (PVD) process of kiwifruit slices and evaluated the optimal drying conditions using artificial neural network (ANN). The results showed that the drying parameters of kiwi fruit were significantly affected by content of vitamin C and drying parameters. The optimal conditions predicted by ANN were 65°C, 12 min for VPD and 2 min for APD. Epstein et al. [8] replaced low-temperature fixed-bed drying with high-temperature drying and studied the impact on the environment. The results showed that the low-temperature fixed-bed drying can reduce greenhouse gas emissions by up to 90% at similar costs. Stefaniak et al. [9] evaluated the environmental impact of three different types of dryers used in sewage treatment plants and found that the drying process was the main cause of energy consumption. The energy consumption of intermittent dryer was 39.6% lower than that of belt dryer. The use of belt dryer and container dryer for sludge treatment will increase the environmental burden. Prosapio et al. [10] comparatively analysed the sensitivity of the combination of traditional freeze-drying and osmotic dehydration combined with freeze-drying for strawberries. The study shows that the emissions of latter are reduced by 25% compared with the traditional freeze-drying process. Changes in the amount of fertilizer, the power of the condenser and vacuum pump have little effect on the results, and the transportation distance has a certain impact on environmental changes.

The purpose of sensitivity analysis is to describe the influence of the input variables of the model on the output of the model [11]. It is mainly divided into local sensitivity analysis (LSA) and global sensitivity analysis (GSA). The LSA considers the output variability caused by the change of the input factor near a certain value [12]. Compared with the LSA, which can only research the relationship around the data points, but cannot consider the interaction between the input variables, the global sensitivity analysis considers the changes of input factors in the whole change space, which can further study the relationship between input and output in the whole input space. Therefore, more and more researches adopt the global sensitivity method. Standard regression coefficient (SRC) method and treed Gaussian process (TGP) are two commonly used sensitivity analysis methods. In the SRC method, the average value of the variable is subtracted from the variable and the result is divided by the standard deviation of its sample. The larger the absolute value of SRC of a variable is, the more important the variable is [13]. This method is fast and easy to understand and can provide intuitive quantitative results for researchers. Moreover, compared with ordinary regression coefficients, SRC removes the unit of independent variables, and their values can directly compare the relative importance of input factors. But the disadvantage of this method is that it is only suitable for a linear model [14]. TGP method is based on Bayesian treed Gaussian process model, combined with tree segmentation model, linear model, stationary Gaussian process and Bayesian analysis. The model can generate a high-fidelity response surface and has a high performance of out of sample prediction and good characteristics of the tree-based method [15]. Using this method to analyze the sensitivity of the environmental impact of the MPSD system can not only find the parameters that dominate the environmental impact, but also investigate the interaction between the parameters. Yang et al. [14] used Morris design, extended Fourier Amplitude Sensitivity Test and TGP methods to perform a global sensitivity analysis on the energy assessment of a retail building. The results show that the TGP method is the best choice for comprehensive performance, whether in terms of accuracy or computational cost. In this paper, SRC and TGP methods are used to analyze the environmental impact of the system. While playing the advantages of the two methods, the reliability of the analysis results is improved.

The innovation of this study is to explore the important parameters in the design and operation of the MPSD system from the perspective of environmental protection for the first time. This can be a useful tool for manufacturers of milk powder and the related equipment to understand the environmental impact of the system on the life cycle and explore opportunities for sustainable improvement, as well as help these manufacturers to determine the priority of actions from the perspective of environmental protection.

In this article, the R software is used to establish the design model of the system, which can output the technical parameters of the system according to the input of capacity demand. The consumption of resources and energy of the whole life cycle of the system are determined by these parameters. Then the LCA model based on matrix is established to quantify the environmental impact of the system. Finally, the sensitivity analysis of the environmental impact of the system is carried out by using TGP and SRC methods, and the parameters that play a leading role in the environmental impact are identified.

2 Materials and methods

2.1 The production process of milk powder

The main equipment of the system (Fig. 1) includes an air heater, drying tower, vibrating fluidized bed, cyclone separator, intake fan and exhaust fan. The pretreated concentrated milk is dispersed into small droplets in the drying tower through the atomization process and is mixed with hot air heated by an air heater at the same time. After the moisture in the droplet is removed for the first time, it enters the vibrating fluidized bed for the second drying, and then is screened and collected. Some hot air containing powder is collected by a cyclone separator. Finally, the exhaust gas is discharged through the exhaust fan.

Fig. 1

Two-stage milk powder spray drying system [16, 17].

2.2 The design model of the system

The research object of this article is a full-fat milk powder spray drying system with an annual output of 3,000 tons. The design model of the system mainly includes the size of the drying tower and fluidized bed, the motor power of the inlet fan and exhaust fan. Due to a large number of formulas, it cannot be listed in this paper. Only four of the more important formulas are listed. The calculation results of the following four formulas directly affect the consumption of resources and energy in the whole life cycle of the system, thereby determining the impact of the system on the environment. $V = \frac{W}{n}$ (1)

The calculation of the effective volume of the drying tower is shown in Equation (1). W is the evaporation of water during the process of drying the concentrated milk into milk powder, n is the intensity of evaporation. The volume of the drying tower V directly affects the total weight of the main equipment of the system, thereby affecting the consumption of stainless steel. $L = \frac{W}{γ_{2} - γ_{1}}$ (2)

From Equation (2), the amount of dry air L can be calculated, where W is the evaporation of water, γ₂ is the relative humidity of the exhaust gas, and γ₁ is the relative humidity of the hot air. $D = \frac{Q}{γ \times η}$ (3)

According to Equation (3), the steam consumption of the air heater can be calculated, where Q is the heat required for heating the air, which is related to the amount of dry air and the enthalpy of hot air, γ is the latent heat of condensation at the current temperature of the steam, and η is the heat utilization coefficient. $N = \frac{q \times H}{1000 \times 3600 \times η_{1} \times η_{2}} \times B$ (4)

From Equation (4), the motor power of the fan can be calculated, where q is the inlet air volume, H is the total air pressure, η₁ is the internal efficiency of the fan, η₂ is the mechanical efficiency of the fan, and B is the safety factor of the motor capacity. The motor power directly affects the power consumption of the system.

2.3 Goal and scope definition

The objective of this research is to quantify the environmental impact of the milk powder spray drying system throughout the life cycle and find out the factors that have a significant impact on the environmental load of the system. The functional unit is a MPSD system with an annual output of 3000t. The system works 12 hours a day, 300 days a year, and the service life of the system is 20 years. As shown in Fig. 2, the system boundary of this study is the entire life cycle of the system, including the acquisition of raw materials, manufacturing and assembly, transportation, use, maintenance and end-of-life.

Fig. 2

System boundary of life cycle assessment of the MPSD system.

2.4 Inventory analysis

This research is organized according to the two national standards formulated by the National Environmental Management Standardization Technical Committee corresponding to the latest international standards: GBT24040-2008 and GBT24044-2008 and based on the Chinese Life Cycle Database (CLCD) in the eBalance software jointly developed by Sichuan University and IKE Environmental Technology Co., Ltd. In addition, the data in related literature are also referenced. The consumption of stainless steel, electricity and steam of the system are calculated by the design model of the system according to the capacity demand. The accuracy of the data is verified by consulting professionals from milk powder manufacturers.

As the raw materials of the equipment of the system are mainly stainless steel, the stage of raw materials mainly refers to the production process of stainless steel. The stage of manufacturing and assembly includes the process of manufacturing and assembling the parts of the system, which mainly consumes electricity and stainless steel. And the stage of transportation the system involves two sections. The first section is in the stage of the manufacturing and assembly, the system is transported from the plant of equipment manufacturing to the plant of milk powder production for installation at this stage. And the second section is in the stage of end-of-life, the system is transported from plant of milk powder production to the recycling site. According to the total weight of the equipment, the medium truck with a carrying capacity of 8t is used in this stage. The stage of use is the process of milk powder production, in which electricity and steam are consumed by the system. The maintenance phase mainly focuses on the cleaning of drying tower and other equipment, which consumes industrial water, acidic and alkaline cleaning agents, and produces a large amount of wastewater. The end-of-life phase includes the recycling and disposal of the main equipment, in which part of the stainless steel is recovered and electricity is consumed.

2.5 LCA model of the system

In order to carry out life cycle assessment of the system, impact assessment methods such as IPCC2007, mid-point CML2001, EDIP97 are used. Considering the main environmental emissions of the system, such as CO₂, SO₂, CO, COD, BOD, etc., the impact categories selected in this study are Global Warming Potential (GWP), Acidification Potential (AP), Eutrophication Potential (Eutrophication Potential, EP), Photochemical Oxidation Potential (POCP). In addition, the Primary Energy Demand (PED) has been added to quantify energy consumption of the system. Finally, The LCA model of the system based on the matrix is established by using R software.

2.6 Sensitivity analysis

The steps of sensitivity analysis in this paper are as follows (Fig. 3): first, according to the relevant literature [1 , 17] on spray drying of milk powder, 9 main parameters of the design and operation of spray drying system are selected, and the interval of parameter changes is determined, as shown in Table 1. By consulting the technical personnel of the milk powder manufacturer, the accuracy and rationality of the range of the relevant parameters have been verified. Then, the Latin hypercube sampling of the parameters in the interval is carried out according to the uniform distribution, and 300 parameters combination samples are generated. The samples are input into the design model and energy consumption model of the system, and the steam consumption, power consumption and the total mass of related equipment of the system are calculated. Finally, SRC and TGP methods are used to analyze the sensitivity of the environmental impact of the system on each parameter.

Fig. 3

Flow chart of sensitivity analysis of the system.

Table 1

Parameters and the ranges of the system

Parameters	Symbol	Unit	Range
Moisture content of the product during primary drying	V1	%	4.5∼6
Moisture content of the product during secondary drying	V2	%	2.5∼3
Inlet air temperature of the drying tower	V3	°C	180∼200
Temperature of concentrated milk	V4	°C	70∼80
Temperature of exhaust gas from drying tower	V5	°C	70∼85
Total air pressure of the inlet fan	V6	mmH2O	80∼160
Total air pressure of the exhaust fan	V7	mmH2O	160∼250
Feed pump speed	V8	rpm	73∼206
Atomization pressure	V9	bar	100∼300

The SRC is a sensitivity analysis method based on linear or monotonic hypothesis in the case of independent factors. The method standardizes the data, and the calculated regression equation is called the standardized regression equation, and the corresponding regression coefficient is the standardized regression coefficient [18]. $Y = b_{0} + \sum_{i = 1}^{n} b_{i}$ (5) ${SRC}_{i} = b_{i} \frac{σ_{i}}{σ_{y}}$ (6)

The calculation of the multiple linear regression model and SRC is shown in Equation (5) and Equation (6), Y is the output of the model, b₀ is the intercept, b_i is the regression coefficient of the i-th parameter, n is the number of parameters, SRC_i is the standard regression coefficient of the i-th parameter, σ_i is the standard deviation of the i-th parameter, and σ_y is the total standard deviation of the model output.

TGP is a global sensitivity method that combines static Gaussian process and decision tree. This method combines TGP with the Sobol method based on variance decomposition. Compared with the SRC method, it can better deal with the nonlinear dynamic models [19]. The method is divided into two steps. Firstly, according to the matrix of input and output variables, the machine learning model of the Gaussian process is obtained, and then the importance degree of different factors is obtained by using the sensitivity analysis method based on variance. This method includes two indicators: main effect S and total effect T, $S_{i} = \frac{Var (E (Z | X_{i}))}{Var (Z)}$ (7) $T_{i} = 1 - \frac{Var (E (Z | X_{- i}))}{Var (Z)}$ (8)

S_i and T_i are the main effect and total effect of the i-th input variable respectively, E is the expected value, Var is the variance value, X_- i is the set of all variables except X_i, Var (E (Z | X_i)) represents the variance of expected values of outputs Z when X_i is specified, and Var (Z) represents the variance of output variable Z. The main effect represents the change in the output caused by the change of each input variable. The total effect also includes the change in the output caused by the interaction between the input variable and other input variables. The higher the main effect or the total effect, the more important the variable is.

3 Results and discussion

3.1 Results of sensitivity analysis

Figure 4 shows the result of sensitivity analysis of the environmental impact of the MPSD system based on the TGP method. The three subgraphs correspond to the main effect, the first-order effect and the total effect of parameters of the system. The graph of main effects shows the change of output (environmental impact) with input (parameters). All inputs and outputs have been standardized to facilitate comparative analysis. The range of horizontal and vertical coordinates is –0.5 to 0.5. It can be seen from Fig. 4-(a) that the impact of the whole life cycle of the system on the environment is significantly positively correlated with the inlet air temperature of the drying tower (V3), the feed pump speed (V8) and the atomization pressure (V9), indicating that the increase of these parameters will significantly improve the impact of the system on the environment. This is because the drying process needs to remove the moisture of the concentrated milk to produce the final powdered product. The higher the inlet air temperature is, the higher the ability of removing water is and the more steam is consumed. The higher the feed pump speed and the atomization pressure, the greater the power consumption. The steam required by the system comes from power plants, while the power structure is mainly coal-fired power plants in China. Electricity production requires a lot of coal, and produces emissions of various substances, such as greenhouse gases. Therefore, the more steam and electricity consumption, the greater the impact on the environment. There is a weak positive correlation between the total air pressure of the exhaust fan (V7) and total air pressure of the inlet fan (V6), which shows that the increase of these variables will also cause the increase of environmental impact, but the impact is small. The results of SRC sensitivity analysis in Fig. 5 can more directly reflect the correlation between parameters and environmental impact.

Fig. 4

Sensitivity of the environmental impact of the system based on the TGP method.

Fig. 5

Sensitivity of the environmental impact of the system based on the SRC method.

Figure 4-(b) and 4-(c) show the first-order effect and the total effect of a variable. The larger the first-order effect and the total effect of a variable, the more important the variable is. The difference between the first-order effect and the total effect is the variance of the output due to the interaction between the variable and other input variables. Therefore, the greater the difference means that the interaction between the variable and others is stronger [20]. It can be seen from Fig. 4-(b) that the inlet air temperature affects the fluctuation of environmental impact by about 82%, the atomization pressure and the feed pump speed account for about 9% and 8% respectively. Figure 4-(c) shows that the atomization pressure and the feed pump speed are about 2–3% lower than the main effect value, indicating that there is a certain inhibitory effect between the variables and others.

According to the results of the sensitivity analysis of the environmental impact on the energy consumption (Fig. 6), it can be found that steam consumption (x3) plays a leading role in the environmental impact. This is further confirmed in Fig. 6-(b), which shows that the steam consumption causes about 70% change of environmental impact and is positively correlated. The effect of power consumption (x2) on the environment is about 29%, and the impact caused by the change of total system weight (x1) is very small. Figure 6-(c) shows that the power consumption is about 1% lower than the main effect value, indicating that there is a certain inhibition effect between this variable and others.

Fig. 6

Sensitivity of the environmental impact to the resource and energy consumption.

To further explore the contribution of the key parameters of the system to the environment, the consumption of resources and energy with the change of the parameters is analyzed. Figure 7 shows the results of the sensitivity analysis of steam consumption of the system based on the TGP method. It can be seen from Fig. 7-(a) that the steam consumption has the strongest correlation with the inlet air temperature that is, the steam consumption increases with the increase of the parameter. Figure 7-(b) shows that inlet air temperature contributes almost 100% to steam consumption. This also explains that the variable plays a dominant role in the total environmental impact, mainly because the steam consumption plays a dominant role in the environmental impact, and the inlet air temperature has the greatest impact on the steam consumption. Figure 7-(c) shows the inlet air temperature decreased by 1–2% compared with the main effect value, which showed that the variable has a certain inhibitory effect on other variables. The results based on SRC sensitivity analysis also confirmed the above findings (Fig. 8).

Fig. 7

Sensitivity of the steam consumption of the system based on the TGP method.

Fig. 8

Sensitivity of the steam consumption of the system based on the SRC method.

Figures 9 10 show the results of the sensitivity analysis of the power consumption. The power consumption has the strongest correlation with the feed pump speed and the atomization pressure, which means the power consumption increases with the increase of the parameters. Figure 9-(b) shows that the feed pump speed and the atomization pressure contribute 40% and 38% to power consumption respectively, which explains why these two parameters have a large contribution to the environmental impact.

Fig. 9

Sensitivity of the power consumption of the system based on the TGP method.

Fig. 10

Sensitivity of the power consumption of the system based on the SRC method.

3.2 The balance between improving the environmental performance and ensuring quality

Although the environmental impact of the system can be significantly reduced by reducing the main parameters of the system, it should be noted that the primary purpose of milk powder production is to ensure the quality of milk powder, which depends on the parameters of the system. The powder properties mentioned here include moisture content, bulk density, hygroscopicity and solubility of the milk powder [21, 22]. The increase of inlet temperature can make the temperature difference between the concentrated milk and the dry air to promote moisture removal [23, 24]. Studies have shown that the bulk density decreases with the increase of the inlet air temperature, which is due to the rapid formation of an airtight film on the droplet surface, and then the formation of bubbles to expand the droplet [25, 26]. The decrease in inlet air temperature can significantly reduce the hygroscopicity, which is due to the lower inlet air temperature will increase the moisture content of the product [27 –29]. The increase of hot air inlet temperature can improve the solubility, which is due to the lower moisture content of milk powder at higher inlet air temperature [30, 31]. In addition, the color of milk powder is also one of the important quality attributes of the product, which largely determines the consumers’ acceptance of the product [32]. The brightness value of milk powder represents the degree of over-drying of the powder. It is found that this value decreases with the increase of air inlet temperature, which may be due to the promotion of the Maillard reaction at a higher temperatures [33]. Therefore, the inlet air temperature not only has a great impact on the environment during the life cycle of the system£¬ but also has a close relationship with the quality of milk powder. Therefore, it is necessary to find the optimal scheme in both the minimizing the environmental impact and ensuring the quality of milk powder when improving the parameters.

3.3 Substitution and reuse of energy

According to the analysis of Fig. 6, the steam and electricity consumption are the dominant factors affecting the environment of the system. Combined with the characteristics of energy structure of China, which is dominated by coal-fired power generation, it can be known that most of the steam consumed comes from coal-fired power plants, which require a large amount of thermal energy. Therefore, this problem can be alleviated in two ways: the reuse of waste-heat resources and looking for alternative energy sources. Since it is challenging to recover heat from the exhaust gas of the MPSD system, the feasibility of recovery can be improved by incorporating the evaporation stage into the recovery scope. However, in the process of milk powder processing, high working temperature and particle load will cause powder deposition in the exhaust heat exchanger, which is the biggest obstacle to heat recovery [34]. Another way to reduce energy consumption is to use renewable energy to replace existing traditional energy sources. At present, many studies have verified the feasibility of using solar energy to replace coal. Camci [35] has carried out a thermodynamic analysis on the spray drying system using solar collector and found that the potential of using renewable energy in the production of milk powder has greatly improved energy efficiency.

4 Conclusions

In this paper, LCA method and global sensitivity analysis method are used to find the parameters that have an important impact on the environment in the MPSD system. The results show that the inlet air temperature of the drying tower has the greatest impact on the environment throughout the life cycle of the system, accounting for about 82%. The atomization pressure and the feed pump speed account for 9% and 8% respectively. The steam consumption caused about 70% of the environmental impact, the power consumption accounting for about 29%, and the change in the total system weight caused the smallest environmental impact. In addition, it is necessary to find the optimal solution while meeting the environmental protection and ensuring the quality of milk powder simultaneously when optimizing parameters such as the inlet air temperature of drying tower. This article not only identifies the factors of the system that have a significant impact on the environment, but also provide manufacturers of milk powder and the related equipment qualitative and quantitative perception of the importance of the factors. Thereby to help them improve the sustainability of the industry more efficiently.

Footnotes

Acknowledgments

We are very grateful to Meng Xing, Xianhao Meng, Xiuling Gao and Ting Cao for their guidance and help in life cycle assessment and sensitivity analysis.

References

, Spray drying technology, Chemical Industry Press (2013), Beijing, China.

Han

, Sun

, Ming

, Chai

and Liao

, Are the Chinese Moving toward a Healthy Diet? Evidence from Macro Data from 1961 to 2017, Int J Environ Res Public Health 17 (2020).

Valsasina

, Pizzol

, Smetana

, Georget

, Mathys

and Heinz

, Life cycle assessment of emerging technologies: The case of milk ultra-high pressure homogenisation, J Clean Prod 142 (2017), 2209–2217.

De Marco

, Riemma

and Iannone

, Uncertainty of Input Parameters and Sensitivity Analysis in Life Cycle Assessment: An Italian Processed Tomato Product, J Clean Prod 177 (2018), 315–325.

Gregory

J.R.

, Montalbo

T.M.

and Kirchain

R.E.

, Analyzing uncertainty in a comparative life cycle assessment of hand drying systems, Int J Life Cycle Ass 18 (2013), 1605–1617.

Chavan

, Vitankar

and Thorat

, CFD modeling and experimental study of solar conduction dryer, Dry Technol (2020), 1–14.

Liu

Z.L.

, Wei

Z.Y.

, Vidyarthi

S.K.

, Pan

, Zielinska

, Deng

L.Z.

, Wang

Q.H.

, Wei

and Xiao

H.W.

, Pulsed vacuum drying of kiwifruit slices and drying process optimization based on artificial neural network, Dry Technol (2020), 1–13.

Epstein

, Lubitz

, Dineen

and Dyck

, Performance evaluation of a non-equilibrium model for low temperature grain drying and simulation of seasonal dryer operation, Dry Technol (2020), 1–17.

Stefaniak

, Zelazna

and Pawlowski

, Environmental assessment of different dewatering and drying methods on the basis of life cycle assessment, Water Sci Technol 69 (2014), 783–788.

10.

Prosapio

, Norton

and De Marco

, Optimization of freeze-drying using a Life Cycle Assessment approach: Strawberries’ case study, J Clean Prod 168 (2017), 1171–1179.

11.

Famiglietti

, Guerci

, Proserpio

, Ravaglia

and Motta

, Development and testing of the Product Environmental Footprint Milk Tool: A comprehensive LCA tool for dairy products, Sci Total Environ 648 (2019), 1614–1626.

12.

Pianosi

, Beven

, Freer

, Hall

J.W.

, Rougier

, Stephenson

D.B.

and Wagener

, Sensitivity analysis of environmental models: A systematic review with practical workflow, Environ Modell Softw 79 (2016), 214–232.

13.

Sharifi

, Ahmadi

, Badfar

and Hosseini

, Modeling and sensitivity analysis of NOx emissions and mechanical efficiency for diesel engine, Environ Sci Pollut R 26 (2019), 25190–25207.

14.

Yang

, Tian

, Cubi

, Meng

, Liu

and Wei

, Comparison of Sensitivity Analysis Methods in Building Energy Assessment, Procedia Engineering 146 (2016), 174–181.

15.

Gramacy

R.B.

and Taddy

, Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with Tgp Version 2, an R Package for Treed Gaussian Process Models, J Stat Softw 33 (2010), 1–48.

16.

, Technology of milk and dairy products, China Measurement Press (2006), Beijing, China.

17.

Feng

, Dairy machinery and equipment, Chinese Light Industry Press (2013), Beijing, China.

18.

Deman

, Kerrou

, Benabderrahmane

and Perrochet

, Sensitivity Analysis of Groundwater Lifetime Expectancy to Hydro-Dispersive Parameters: The Case of ANDRA Meuse/Haute-Marne Site, Reliab Eng Syst Saf 134 (2015), 276–286.

19.

Song

, Tian

, Xu

, Wang

and Li

, Thermal performance of a novel ultrasonic evaporator based on machine learning algorithms, Appl Therm Eng 148 (2019), 438–446.

20.

Zhu

and Chen

, Sensitivity Analysis on Borehole Thermal Energy Storage under Intermittent Operation Mode, Energy Procedia 158 (2019), 4655–4663.

21.

Bansal

, Sharma

H.K.

and Nanda

, Optimisation of Spray Drying Process Parameters for Low-Fat Honey-Based Milk Powder with Antioxidant Activity, Int J Food Sci Technol 49 (2014), 1196–1202.

22.

Nguyen

D.Q.

, Nguyen

T.H.

, Mounir

and Allaf

, Effect of feed concentration and inlet air temperature on the properties of soymilk powder obtained by spray drying, Dry Technol 36 (2017), 817–829.

23.

Islam Shishir

M.R.

, Taip

F.S.

, Aziz

N.A.

, Talib

R.A.

and Hossain Sarker

M.S.

, Optimization of spray drying parameters for pink guava powder using RSM, Food Sci Biotechnol 25 (2016), 461–468.

24.

Tonon

R.V.

, Brabet

and Hubinger

M.D.

, Influence of process conditions on the physicochemical properties of açai (Euterpe oleraceae Mart.) powder produced by spray drying, J Food Eng 88 (2008), 411–418.

25.

Fazaeli

, Emam-Djomeh

, Kalbasi

and Omid

, Effect of spray drying conditions and feed composition on the physical properties of black mulberry juice powder, Food Bioprod Process 90 (2012), 667–675.

26.

Fazaeli

, Emam-Djomeh

and Yarmand

M.S.

, Influence of Black Mulberry Juice Addition and Spray Drying Conditions on Some Physical Properties of Ice Cream Powder, Int J Food Eng 12 (2016), 277–285.

27.

Bakar

, Ee

S.C.

, Muhammad

, Hashim

D.M.

and Adzahan

, Spray-Drying Optimization for Red Pitaya Peel (Hylocereus polyrhizus), Food Bioprocess Tech 6 (2012), 1332–1342.

28.

Juarez-Enriquez

, Olivas

G.I.

, Zamudio-Flores

P.B.

, Ortega-Rivas

, Perez-Vega

and Sepulveda

D.R.

, Effect of water content on the flowability of hygroscopic powders, J Food Eng 205 (2017), 12–17.

29.

Koç

, Güngör

Ö.

, Zungur

, Yalçın

, Selek

İ.

, Ertekin

F.K.

and Ötles

, Microencapsulation of Extra Virgin Olive Oil by Spray Drying: Effect of Wall Materials Composition, Process Conditions, and Emulsification Method, Food Bioprocess Tech 8 (2014), 301–318.

30.

Goula

A.M.

and Adamopoulos

K.G.

, Effect of Maltodextrin Addition during Spray Drying of Tomato Pulp in Dehumidified Air: II. Powder Properties, Dry Technol 26 (2008), 726–737.

31.

Thakur

, Pant

, Naik

R.R.

and Nanda

, Optimization of spray drying operating conditions for production of functional milk powder encapsulating bee pollen, Dry Technol (2020), 1–14.

32.

Zungur Bastioglu

, Tomruk

, Koc

and Ertekin

F.K.

, Spray dried melon seed milk powder: physical, rheological and sensory properties, J Food Sci Technol 53 (2016), 2396–404.

33.

Selvamuthukumaran

and Shukla

S.S.

, Optimization of spray drying conditions for production of bifidus milk powder from cow milk, J Food Quality 29 (2006), 305–318.

34.

Atkins

M.J.

, Walmsley

M.R.

and Neale

J.R.

, Integrating Heat Recovery from Milk Powder Spray Dryer Exhausts in the Dairy Industry, Appl Therm Eng 31 (2011), 2101–2106.

35.

Camci

, Thermodynamic Analysis of a Novel Integration of a Spray Dryer and Solar Collectors: A Case Study of a Milk Powder Drying System, Dry Technol 38 (2020), 350–360.