Process Integration and Optimization for Sustainability

https://doi.org/10.1007/s41660-023-00382-3

ORIGINAL RESEARCH PAPER

A Multi‑objective Mixed‑Integer Linear Model for Sustainable Dairy

Supply Chain with Food Waste and Environmental Pollutants
Iwan Vanany1 · Ivan Darma Wangsa1 · Nofariza Aulia Jeremi1

Received: 15 August 2023 / Revised: 18 October 2023 / Accepted: 28 October 2023

© The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2023

Abstract
This study contributes to dairy supply chain due to the lack of food waste and environmental pollutants. The objective of this
study is to minimize the total system cost and reduce carbon emissions and food waste for a dairy supply chain with multi-
echelons. The system consists of multiple farmers, single processing plant, multi-distributor, and multi-retail customer. This
study presents a multi-objective mixed-integer linear programming (MOMILP). A MOMILP proposed model was validated
to provide insights into dairy industries. A real case study of dairy milk problems in West Java (Indonesia) was solved by
using optimization software. The optimal results show that applying the proposed model can minimize total costs, reduce
food waste, and minimize environmental pollutants. Finally, the results of sensitivity analysis show that the total supply chain
costs and food waste costs are significantly influenced by variance of temperature in fresh milk storage.

Keywords MILP · Multi-objective · Dairy supply chain · Product waste · Environmental pollutant

Introduction the food and beverage industry grew by 1.58% in 2020,

while the non-oil and gas industry declined by − 2.52% in
The food and beverage sector is the largest energy consumer, 2020 (Indrayani 2021). With the growth of the food and
producing 67% of carbon emissions (European Commission beverage industry in both “normal” and pandemic times, the
2015). Dairy products, one of the essential products in the problem of carbon emissions and waste has also increased
food and beverage industry, require constant refrigeration and become a concern of the Indonesian government. For
along the supply chain. This causes the amount of carbon food waste, it is also a concern for the Indonesian govern-
emissions produced in the production, logistics, and dis- ment because it is the largest food waste-producing country
tribution processes to be high and impact the environment in Southeast Asia, with the total amount of food waste in
(Glover et al. 2014; Validi et al. 2014). In addition, dairy Indonesia 20.93 million tons per year (Naurah 2022). The
products also produce food waste, significantly cooling amount of food waste in Indonesia varied from 23 to 48
temperatures which are not constant and fluctuate highly million tons annually between 2000 and 2019 (BAPPENAS
(Mercier et al. 2017). 2021). Carbon emissions are also generated from food waste,
In Indonesia, the food and beverage industry occupies although insignificant compared to transportation and refrig-
an essential position in the 5 years 2015–2019, growing by eration from food and beverage supply chain activities.
8.16% and exceeding the growth of other manufacturing The supply chain of dairy products in Indonesia generates
industries. During the COVID-19 pandemic in 2020–2022, a high amount of carbon emission along the supply chain
with dairy farming, dairy processing plants, dairy product
* Iwan Vanany distributors, and retail stores. In dairy farming, due to the
[email protected] high increase in consumer demand, carbon emissions gener-
Ivan Darma Wangsa ated as much as 1.7 million in 2005 and increased by 18%
[email protected] per year (IDF 2015). In dairy processing plants, there are
Nofariza Aulia Jeremi about 84 dairy processing companies that produce dairy
[email protected] products (Kompas 2022). Almost all of them require refrig-
eration, transportation from suppliers, and distribution of
1
Department of Industrial and Systems Engineering, Institut dairy products and other machinery and equipment, which
Teknologi Sepuluh Nopember, Surabaya, Indonesia

13
Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

are still fuelled by oil. The distributors of dairy products and and production activities processing raw milk (raw milk).
modes of transportation for the distribution of dairy products In research, Wangsa et al. (2023) developed an integrated
throughout Indonesia still use gasoline and diesel-fuelled optimization model considering carbon taxes and food waste
ships and trucks with long transportation distances due to in the food supply chain with the object of research on agri-
the vastness of Indonesia. In addition, cold storage for dairy cultural products. So, what distinguishes it is the addition
products is also needed to support sales and distribution in of the objective function of calculating food waste to the
the Indonesian market. In the retail sector, refrigeration is objective function.
also needed to store dairy products to ensure that the prod- Mogale et al. (2017) proposed a mixed integer non-linear
ucts do not spoil quickly and remain fresh. programming (MINLP) programming model developed after
The remainder of this paper is organized as follows: studying Indian wheat supply chain scenarios, where the
“Research Gap” discusses previous studies and finding the aim was to minimize the transportation, storage, and opera-
research gap. “Research Methods” presents the design of tional costs of food grains issued for the efficient transfer
a MILP formula to optimize the dairy case study. “Practi- of grain from producing countries to consuming countries.
cal Implementation” implements the model with a real case In contrast, Li et al. (2019) determined minimizing deliv-
study in Indonesia. A discussion and sensitivity analysis are ery time as the objective function in the MILP model. The
conducted in “Sensitivity Analysis and Managerial Insight.” objective function of minimizing the total cost of distribu-
Finally, concluding remarks and future directions are pre- tion is indeed widely used in several journals; in research,
sented in “Conclusions and Future Directions”. Nasr et al. (2022), the objective function has paid attention
to disposal costs in the primary model, but it is different
from research (Patidar and Agrawal 2020) which compares
Research Gap two scenarios such as with and without paying attention to
the perishable condition of the product to calculate the cost
Research and modeling on food supply chain (FSC) have when the product has decayed. In contrast, two studies that
received much attention from previous researchers. Until focused on the objective function of cost maximization made
now, many studies have solved problems using optimiza- a model to calculate profit maximization, which can be cat-
tion models. The method often used in optimization models egorized as primary and secondary profits, reducing some
aimed at solving problems in various studies according to costs (de Keizer et al. 2017).
actual conditions is the MILP method developed accord- Similarly, aiming to maximize the model, Vanzetti et al.
ing to the research case study. The MILP method handles (2018) model to maximize profit (benefit) by considering
mathematical optimization problems involving discrete and revenue from sales and costs of raw materials, production
continuous variables. Both the objective function and the costs, and setup costs. In research, Chaabane et al. (2022)
constraints are linear. The MILP is an optimal method for focused on models with three objectives, namely minimizing
solving many real problems, as discussed in many agricul- total costs related to logistics, considering GHG emissions
tural and livestock industry studies. The MILP model was generated by production and distribution activities, and min-
developed in the fishing industry to solve production–distri- imizing job instability in various manufacturing locations.
bution network problems. Handayani et al. (2021) developed Meanwhile, research by Wang et al. (2021) created a model
the MILP model by combining traceability and carbon emis- to optimize costs, delivery times, transportation emissions,
sions in production distribution problems that have yet to be and shipping during high-demand time windows. In addi-
discussed simultaneously in previous studies. A develop- tion, the same research was done by Purnomo et al. (2022).
ment model is obtained with the objective function of mini- The authors conducted a study on minimizing total costs by
mizing the total cost of ten cost components with limitations considering carbon emissions in the fish closed-loop supply
on-demand fulfillment, production level, balance quantity of chain problem case study.
raw fish with canned fish products, and achievement of total A mathematical model for a three-echelon agri-food sup-
emissions with emission threshold values. ply chain network was proposed by Goodarzian et al. (2022).
Moreno-Camacho et al. (2023) also developed the MILP The model takes into account carbon emissions, coefficient
model with problems in the livestock industry, namely, water, and time window. The authors proposed a bi-objective
the object of cow’s milk. The developed model considers MINLP to reduce total cost and C ­ O2 emissions. Due to its
the reduction of GHG emissions and the improvement of lengthy computing time, the augmented Epsilon-constraint
social health and living conditions. Like Handayani et al. approach was utilized for small- and medium-sized problems
(2021)’s research, in this study, carbon emissions were con- but could be more effective for large-scale problems. The
sidered only due to transportation activities between points model was solved using a hybrid meta-heuristic approach

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

known as Hybrid Multi-Objective Ant Colony Optimiza- and financial aspects was developed by Gharye Mirzaei
tion with Multi-Objective Simulated Annealing (HACO- et al. (2023). The paper investigated a MILP model to
SA). The results demonstrated the superiority of HACO- improve economic objectives under uncertainty, consid-
SA over the original methods in medium- and large-sized ering weather conditions and economic fluctuations. A
issues, while the Augmented Epsilon-constraint technique real-world case study of dates farms, which substantially
performs better in small- and large-sized problems. Sus- impact GDP, job creation, export, and packing, attests
tainable transportation is crucial for reducing greenhouse to the model’s effectiveness. With an accuracy of 1.9 to
gas emissions, especially in cities like Tehran. Dorcheh and 2.8%, the hybrid method—which is based on the Whale
Rahbari (2023)’s study aimed to determine optimal slaugh- Optimization method (WOA), particle swarm optimization
terhouse capacity, identify the best transport routes, and (PSO), and a hybrid algorithm (WOA-PSO)—is useful and
reduce vehicle usage, transportation costs, and greenhouse practical for resolving large-scale issues. Khazaeli et al.
gas emissions. The model was presented in two phases using (2023) developed a mathematical model for an agricultural
the General Algebraic Modeling System (GAMS) software. supply chain by considering product quality. The study
According to the findings, the model could cut greenhouse aims to improve quality and profit by optimizing a multi-
gas emissions costs by 17.29%, 22.82%, and 23.08% and echelon cold chain for perishable agricultural products.
transportation costs by 0.3% and 9.8%. A MILP mathemati- The cold chain consists of four echelon supply chains:
cal model considering a closed-loop supply chain was devel- supply, processing, storage, and customer. The authors
oped by Rajabi-Kafshgar et al. (2023). The study aimed to integrated decision-making in facility location, assigning
reduce the supply chain cost in a network of agricultural nodes, vehicle allocation, processing selection, and refrig-
supply chains. The model utilized metaheuristic and hybrid erated facility temperature settings by using a bi-objective
algorithms to enhance the intensification and diversification MINLP. The model preserves 93% of the quality of the
phases. The quality of these algorithms was investigated and original crop while making a 29% profit against a total
compared. cost of 71, indicating a 7% quality loss. Six million dollars
A robust optimization for the agri-food supply chain dur- were invested in the supply chain, going toward refrig-
ing the COVID-19 pandemic crisis was developed by Rah- eration, establishing facilities, acquiring high-speed vehi-
bari et al. (2023a). This study aimed to develop a model of cles, and acquiring raw materials. Yuniarti et al. (2023)
the canned food supply chain under uncertain conditions to proposed a MILP model for a closed-loop supply chain
analyze it strategically during the pandemic. Robust opti- considering carbon emission and product traceability. The
mization was used to address uncertainty and determine the study established a model for production distribution and
best strategy. The findings indicated that expanding canned multi-period food waste recycling. The findings show that
food exports to neighboring countries with economic justi- by trading-off production and distribution rates with food
fication was the best strategy, reducing supply chain costs, waste, the model can reduce carbon emissions while mini-
increasing human resources, and maximizing vehicle capac- mizing the total costs.
ity and production throughput. The analysis has significantly Research on mathematical modeling in the FSC has
impacted the agri-food industry, causing financial losses and received much attention from previous researchers. There
pressure on companies. Rahbari et al. (2023b) investigated have been many studies to create mathematical models to
a location-inventory-routing problem for a multi-echelon, solve problems in FSC. From the review of previous research
multi-period, multi-product, and multi-objective to study the (Table 1), the Mixed Integer Linear Programming (MILP)
canned food supply networks under uncertain conditions. method is the most famous mathematical method used to be
The model tried to maximize job opportunities while mini- developed by considering the problems of the agriculture
mizing costs and consumer dissatisfaction. Strategies such and livestock industries. Quite a lot also involves multi-
as carbon caps and carbon trade are used to reduce environ- echelons, products, and periods. To the best of the authors’
mental harm. Robust fuzzy stochastic programming (RFSP) knowledge, no one still makes a mathematical model for
was used to deal with uncertainty. According to the study, the dairy supply chain with multi-echelons, products, and
the RFSP strategy had a more significant impact on objective periods. In addition, there still needs to be a mathemati-
functions. Sensitivity analysis showed considerable impacts cal model for the dairy supply chain and simultaneously
on farm purchase costs and selling prices to overseas clients. consider the total cost, food waste, and environment in the
A mathematical model for the optimization of the problem. Overall, the main contributions of this paper are
agricultural supply chain by considering environmental outlined as follows:

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Table 1  Summarizes recent studies in the previous food supply chain (FSC) literature
Author(s), year Echelon Products Period Pollutant Type of products Objective function Model Solution approach

13
Mogale et al. (2017) Multi Single Multi NA Food grain Min total costs MINLP Metaheuristic — chemi-
cal reaction optimization
(CRO) algorithm
de Keizer et al. (2017) Multi Multi Multi NA Fresh agricultural Max profits under quality constraints MILP Exact algorithm — CPLEX
12.6
Vanzetti et al. (2018) Single Multi Multi NA Sawmills Max net profit MILP Exact algorithm — CPLEX
Li et al. (2019) Multi Multi Single NA Perishable Min total cost and delivery time MILP Exact algorithm — branch
and bound algorithm
Hossain and Jahan (2019) Multi Single Single NA Rice Min the total cost of transport MILP Exact algorithm — CPLEX
Patidar and Agrawal (2020) Multi Multi Multi NA Agricultural Min total cost of distribution MILP Exact algorithm — LINGO
17.0
Bacchetti et al. (2021) Multi Multi Multi NA Fresh and frozen Min total costs MILP Exact algorithm — CPLEX
Mosallanezhad et al. (2021) Multi Single Single NA Shrimp Min total costs MILP Metaheuristic — Genetic
Algorithm (GA), simulated
annealing (SA), and Kesh-
tel algorithm (KA)
Esteso et al. (2021) Multi Multi Multi Yes Fresh fruits vegeta- Max net profit MILP Exact algorithm — Gurobi
bles 8.0.1. solver
Wang et al. (2021) Multi Multi Multi NA Fresh fruits Min total cost, delivery time, emis- MIP Exact algorithm — CPLEX
sions 12.6.2
Nasr et al. (2022) Multi Multi Multi NA Agricultural Min total distribution costs MILP Exact algorithm — CPLEX
Handayani et al. (2021) Multi Multi Multi Yes Canned fish Min total cost MILP Exact algorithm — Solver
Purnomo et al. (2022) Multi Multi Multi Yes Fresh & waste fish Min total cost MILP Exact algorithm — Open-
Solver 2.9.3
Goodarzian et al (2022) Multi Multi Multi Yes Citrus Min total costs, emission MINLP Metaheuristic — Hybrid
multi-objective ant colony
optimisation with multi-
objective simulated anneal-
ing (HACO-SA)
Chaabane et al. (2022) Multi Multi Multi Yes Frozen food Min logistics cost and emission; max MINLP Exact algorithm — weighted
number of employees sum method — CPLEX
Rajabi-Kafshgar et al (2023) Multi Multi Single No Pistachio Min total cost MILP Metaheuristic — genetic
algorithm (GA), imperial-
ist competitive algorithm

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

(ICA), Keshtel algorithm
(KA), social engineering
optimizer (SEO)
Rahbari et al. (2023a) Multi Multi Multi No Canned food Min total cost MILP Exact algorithm — Torabi
and Hassini (TH) method
Process Integration and Optimization for Sustainability
Process Integration and Optimization for Sustainability

• Developing a mathematical model formulation, the

sum method — OpenSolver

sum method — OpenSolver

Exact algorithm — weighted

Exact algorithm — weighted

hybrid algorithm based on
optimization (PSO), and a

Exact algorithm — CPLEX

objective function, and constraints representation for the

Exact algorithm — Solver

(WOA), particle swarm
optimization algorithm
Metaheuristic — whale
optimal assignment of the dairy supply chain

them (WOA–PSO)
• Formulating a multi-objective model to minimize total
Solution approach

costs, total waste, and total environmental pollutants

simultaneously
• Developing a weighted sum method to find the optimal

2.9.3

2.9.3
solutions and trade-offs between their objectives
• The proposed model is implemented in a real case study
in Indonesia and offers several managerial implications
for the dairy industry
Model

MILP
MILP

MILP
MILP

MILP
MILP

Min total costs, pollutants, food

Min total costs, emissions, food

Research Methods
Min total costs and emission
Max total profit and quality

Description of System
Objective function

The proposed research model uses the MILP model, which

Min total cost
Min total cost

aims to determine the value of the decision variable that

is set either an unknown amount (decision variable) or an
waste

waste

optimized decision (objective function). The MILP model

is used in several studies as a method of solving real prob-
lems that can provide optimal results, one of which is in
Fresh agricultural
Type of products

Moreno-Camacho et al. (2023)’s research which takes into

Dairy product
Dairy product

account carbon emissions in his research by categorizing

Agricultural
Agricultural
Date farms

them into two types, namely, carbon emissions in trans-

portation activities and emissions in production activities,
processing raw milk into finished products. For the objec-
tive function of food waste, the primary reference model is
Pollutant

Wangsa et al. (2023)’s research on fresh food supply chain

systems with multiple farmers, a single processor, multi-
Yes

Yes
Yes
Yes
No

No

distributors, consumers, and multi-periods. The model

in this study optimizes total purchasing, inspection, food
Period

Single

waste, packaging, cold storage, transportation, and car-

Multi

Multi
Multi

Multi
Multi

bon emission costs by optimizing warehouse inventory

and product delivery.
This study was conducted on the supply chain activi-
Products

Multi

Multi
Multi

Multi
Multi
Multi

ties for producing fresh cow’s milk in Bogor, West Java,

Indonesia, as shown in Fig. 1. The series of activities
carried out were the collection of fresh cow’s milk by
Echelon

the supplier and being a supplier for the milk production

Multi

Multi

Multi
Multi

Multi
Multi

factory, then distribution activities. Fresh bowls of milk

are distributed from suppliers to the production factory,
packaging production activities, storage in warehouses to
Gharye Mirzaei et al. (2023)

maintain temperature and quality, and delivery of finished

Moreno-Camacho et al.

products to consumers. This research focuses on five milk

Khazaeli et al. (2023)
Yuniarti et al. (2023)

Wangsa et al. (2023)

Table 1  (continued)

suppliers located in the West Java region (i = 1, 2, …,

Author(s), year

5), which are located in several cities, namely, Bogor,

This research

Sukabumi, Bandung, and Garut, which sell their fresh

(2023)

milk to the production plant. Then, the production factory

processes and packages fresh milk as raw materials into

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

Fig. 1  The observed system in this study

two finished products (p = 1, 2), namely, fresh milk and milk will be rejected and disposed of as food waste at
UHT (ultra-high temperature) milk. The finished product the cost of deterioration. Products are delivered to retail
in packaged milk will be sent to the distribution center/ consumers on the same day. Other operational issues also
DC (j = 1, 2) located in the Sentul and Cisarua areas. It refer to the inspection, storage, and distribution of carbon
will be distributed to retail supermarkets (k = 1, 2, …, 5) footprint. The flow of the production process is shown in
with three types of supermarkets, namely, hypermarkets, Fig. 2.
supermarkets, and corporate supermarkets. The stage after conducting a literature study regarding
The activity of purchasing milk from suppliers to dis- research and in previous research is the next stage of data
tributing finished milk to final consumers is carried out collection. Direct interviews with dairy product factory
for 7 days. Dairy products require refrigeration (cold stor- practitioners in the West Java region collected data. In
age) to maintain the temperature so that the products are the interview, information was obtained regarding the
not damaged with a temperature requirement of 4–7 °C, existing conditions, parameters, and variables related to
with a warehouse capacity of 2,000,000 l for each dairy this study. In addition, there is a discussion regarding
product. The distance between the processing production research limitations and assumptions, first in a West Java
plant and DC 1 and 2 is 8.5 km and 27.4 km, respectively. region consisting of a production factory and multiple
Each distance must be covered in distribution activities DCs. Customer demand rate, production rate, and defect
from DC 1 to hypermarket A is 7.2 km, hypermarket B is rate are uniform. The three planning time horizons or
13.4 km, company supermarket is 24.3 km, supermarket limited periods, namely, within 1 month, with process-
A is 20.1 km, and supermarket B is 24 km. While the dis- ing from upstream to downstream taking 7 days. The
tance that must be taken from DC 2 to hypermarket A is four dairy products studied were subjected to quality
23.9 km, hypermarket B is 26.4 km, corporate supermarket control/QC to check their quality before and after being
is 4.2 km, supermarket A is 31.2 km, and supermarket B is packaged or at the end of the production process at the
28.1 km. In this study, suppliers sent fresh milk, and QC factory. The five packaged products are delivered to the
inspectors inspected and sorted the products. If milk does distribution centers by carriers less than LTL (less than
not meet the company’s operational quality standards, the truckload) trucks and distributed to several customers

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

Fig. 2  The flow of the dairy supply chain system

by truck. In this study, all deliveries will be by truck to as follows. In this study, we proposed a weighted sum
the supermarket industry, hypermarkets, and supermar- method to find the optimal solutions to these objectives.
kets. The six determinations of cooling temperatures are In order to develop the problem mathematically, indi-
assumed to be the same daily for all products. Finally, ces, parameters, and decision variables are shown in the
carbon emissions come from inspection, packaging, and appendix. Based on the descriptions, the proposed model
storage, which require refrigeration because dairy prod- is as follows: Eq. (1) explains that there is a weighting
ucts are perishable and transported. method to minimize total costs, total waste, and total
environmental pollutants, which are carried out in these
Model Development objective functions.

(1)
( ) ( ) ( )
Furthermore, this subsection describes developing the Minimize Z = 𝛼 x Z1 + 𝛽 x Z2 + 𝛾 x Z3
proposed model. To develop a mathematical model,
Equation (2) shows that the objective of the total cost has
input data is collected as parameters at the data collection
been modified according to the conditions in the case study of
stage. To solve the numerical experimental model, it was
a dairy product processing company in the West Java Region,
carried out with OpenSolver 2.9.3 software because of its
and the research focuses on the distribution process. There are
effectiveness in solving the large problem. The result of
six-ten cost components that must be minimized; these costs are
the calculation of the mathematical model is the optimal
the cost of buying milk from the supplier; the QC cost of milk
total cost, total waste, and total environmental pollut-
from the supplier at the production factory; the cost of milk
ants. In this study, the amount of weight determined by
deterioration at the production factory; the cost of production,
the deterministic has been determined by the case study
and packaging of milk; the cost of cooling milk at the produc-
dairy product practitioner with priority for cost (α = 60%),
tion factory before inspection; production factory cooling costs
objective function of waste (β = 30%), and environmental
after inspection; storage costs in a production factory, distribu-
pollutant (γ = 10%). Total costs include purchasing costs,
tion costs from production plants to DC; quality costs at DCs;
inspection costs, packaging costs, production costs, trans-
milk deterioration costs in DC; milk product cooling costs in
portation costs, the cost of waste, and inventory holding
DC; storage costs in DC; transportation costs to consumers;
costs. Based on the parameters and assumptions needed
and pollutant costs for transportation, cold storage, and waste
by the model, the MILP problem can be formulated
activities:

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

T I P T I P T I P T P
∑ ∑∑ ∑ ∑∑ ∑ ∑∑ ∑ ∑
Z1 = CRtip ⋅ ukudtip + CStip ⋅ zkudtip + CWKUDtip ⋅ ekudtip + CPtp ⋅ xxtp
t=1 i=1 p=1 t=1 i=1 p=1 t=1 i=1 p=1 t=1 p=1
T I P
∑∑∑
+ CIRtip ⋅ irtip ⋅ COPp
t=1 i=1 p=1

P T P
∑T ∑ ∑ ∑
+ t=1
CIPtp ⋅ iptp + CHPtp ⋅ iptp
p=1 t=1 p=1

T J P T J P T J P
∑ ∑∑ ∑ ∑∑ ∑ ∑∑
+ CDtjp ⋅ ddctjp ⋅ Sj + CK tjp ⋅ wdctjp + CWDCtjp ⋅ edctjp
t=1 j=1 p=1 t=1 j=1 p=1 t=1 j=1 p=1
(2)
T J P T J P T J K P
∑ ∑∑ ∑ ∑∑ ∑ ∑∑ ∑
+ CIDtjp ⋅ idtjp ⋅ COPdctjp + CHDCtjp ⋅ idtjp + CT tjkp ⋅ qtjkp ⋅ Rjk
t=1 j=1 p=1 t=1 j=1 p=1 t=1 j=1 k=1 p=1

T P T P T P
∑ ∑ ∑ ∑ ∑ ∑
+ CET tp ⋅ tettp + CEI tp ⋅ teitp + CEDtp ⋅ tedtp
t=1 p=1 t=1 p=1 t=1 p=1

Equation (3) defines the objective function of minimizing irtip = ir(t−1)ip + ukudtip − zkudtip − ekudtip ;∀p, ∀i, ∀t (7)
total waste in the two stages of the QC processes of dairy prod-
ucts in production factories and DCs. Equation (4) represents J
∑
the objective function of minimizing total environmental pol- iptp = ip(t−1)p + xxtp − ddctjp ≤ Xtp ;∀p, ∀t (8)
lutants in the distribution, cooling, and food waste activities: j=1

T I P T J P

(3) ddctjp = wdctjp ;∀p, ∀j, ∀t (9)

∑ ∑ ∑ ∑ ∑ ∑
Z2 = DRKUDtip (%) ⋅ zkudtip + DRDCtjp (%) ⋅ ddctjp
t=1 i=1 p=1 t=1 j=1 p=1

Equations (10) and (11) show the supply of dairy products

T
∑ J
∑ K P
∑ ∑[ at the DCs. Equation (12) ensures the level of milk supply after
QC at the DCs plus the number of milk units that pass QC at
]
Z3 = (ddctip ⋅ Sj + qtjkp ⋅ Rjk ) ⋅ ET tjkp
t=1 j=1 k=1 p=1 the DCs plus the amount of food waste in DC must be greater
T
∑ I
∑ J
∑ P
∑ than the number of units sold by consumers:
(4)
[ ]
+ (irtip + iptp + idtjp ) ⋅ EI tijp
K
t=1 i=1 j=1 p=1 ∑
T I J P idtjp = id(t−1)jp + wdctjp − edctjp − qtjkp ≤ Btjp ;∀p, ∀j, ∀t
∑ ∑ ∑ ∑ [ ] k=1
+ (ekudtip + edctjp ) ⋅ EDtijp (10)
t=1 i=1 j=1 p=1
J
Subject to:
∑
iptp + xxtp ≥ ddctjp ;∀p, ∀t (11)
Equations (5) and (6) ensure that the total milk supplied j=1
by the milk supplier and purchased by the production factory
must be equal to the production capacity: K
∑
idtjp + wdctjp + edctjp ≥ qtjkp ;∀p, ∀j, ∀t (12)
I I I
∑ ∑ ∑ k=1
zkudtip + ekudtip ≤ ukudtip ;∀p, ∀t (5)
i=1 i=1 i=1 Equation (13) shows the fulfillment of demand. The
amount of milk sold to consumers has the same value as the
I
∑ greater demand:
zkudtip = xxtp ≤ ytp ⋅ PCRtp ;∀p, ∀t (6)
i=1 qtpjk ≥ dtpjk ;∀p, ∀j, ∀k, ∀t (13)
Equations (7) and (8) guarantee stock availability at pro- Equation (14) ensures that the total production and transpor-
duction plants. Equation (9) shows the accuracy of the milk tation pollutants generated through the distribution stage, prod-
distributed from production factories to DCs and milk products uct cooling (cold storage), and waste for dairy products accord-
that have been QCs: ing to the planning time must not exceed the threshold limit:

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

T J K P �
∑ ∑∑ ∑
(ddctjp ⋅ Sj + qtjkp ⋅ Rjk ) ⋅ ET tjkp
� factory and two distribution centers. The level of customer
t=1 j=1 k=1 p=1
T I J P �
demand, production levels, and defect rates is uniform.
This research focused on UHT (ultra-high temperature)
∑ ∑∑∑ �
+ (irtip + iptp + idtjp ) ⋅ EI tijp
t=1 i=1 j=1 p=1
T I J P � T milk with four types of milk product packaging, namely,
TELV t
∑∑∑∑ � ∑
+ (ekudtip + edctjp ) ⋅ EDtijp ≤
t=1 i=1 j=1 p=1 t=1 65-ml bottle packaging, 125-ml carton packaging, 250-ml
(14) carton packaging, and 1-l carton packaging. The product
Equations (15–17) ensure that the amount sent to custom- distribution covers West Java and surrounding areas. Data
ers is within the capacity of the trucks used in the distribu- related to consumer retail, such as data on demand for
tion process: dairy products by retail consumers, distance data from DC
� ∑P � to customers, transportation costs, and environmental pol-
p=1
ukudtip lutants for distribution activities. Dairy products in this
≤ ylti ;∀i, ∀t (15) study were inspected and, after that, sent to consumers
L
according to demand by using trucks with a capacity of
� ∑P � 12,000 l.
p=1
ddctjp
≤ ydtj ;∀j, ∀t (16)
L
Results and Analysis
� ∑P
After developing the model, optimal global results are
�
p=1
qtjkp
≤ yctjk ;∀j, ∀k (17) obtained. The results of the numerical calculations are
L
obtained after running the model on optimization software.
Equation (18) shows that the decision variable is non- The results show that the three objective functions have
negative and has an integer value:
zkudtip , ekudtip , ukudtip , xxtp , ddctjp , edctjp , irtip , iptp , wdctjp , qtjkp , idtjp ≥ 0 (18) Table 2  Data on production factories and DCs
Parameter Value UoM
Linearization PCRtp U [114,000; 684,000] Liters
DRKUDtip U [1.5%; 4.2%] –
According to the non-linearity of the proposed objective DRDCtjp U [2.2%; 4.0%] –
functions, Eq. (1), the variable ψ and Eq. (19) are intro- CRtip U [6,200; 13,200] IDR/liter
duced for linearization purposes: CStip U [100; 700] IDR/liter
( ) ( ) ( ) CWKUDtip U [100; 150] IDR/liter
̂1
Z ̂2
Z ̂3
Z CWDCtjp U [85; 340] IDR/liter
Minimize 𝜓 = 𝛼 +𝛽 +𝛾 (19)
Z1 Z2 Z3 CPtp U [3,447; 7,583] IDR/liter
CIRtip U [430; 1,230] IDR/liter
Subject to. Equations (5–18). CIPtp U [800; 1,700] IDR/liter
Where the variables of Z ̂1, Z
̂2 , and Z
̂3 denote the initial CDtjp U [80; 110] IDR/liter/km
values of the first, second, and third objectives, respec- CK tjp U [100; 1,200] IDR/liter
tively; and α; β; γ are the weighting coefficients and their CHPtp U [1,190; 3,800] IDR/liter
sum equals 1. CIDtjp U [430; 1,530] IDR/liter
CHDCtjp U [800; 1,900] IDR/liter
Sj U [8.5; 27] Km
CET tp 100 Liters.CO2-eq
Practical Implementation CEI tp 100 Liters.CO2-eq
CEDtp 100 Liters.CO2-eq
Data TELV t 85,000 Liters.CO2-eq
ET tp 0.008 Liters.CO2-eq/lt
Tables 2 and 3 present data for a real-life case study of a EI tp 0.006 Liters.CO2-eq/lt
dairy milk company located in Bogor, West Java, Indone- EDtp 0.005 Liters.CO2-eq/lt
sia. The company has several markets in various regions, Xtp 2,000,000 Liters
but in the case study, this research only examines one Btjp 1,000,000; 1,500,000 Liters
region in West Java, which consists of one production L 12,000 Liters

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

Table 3  Data for consumer Parameter Consumer Value UoM

demand, distance, emission
pollutants, and cost Dtkp ;d = 1 Hypermarket—A U [67,905; 75,4650] Liters
Hypermarket—B U [68,085; 75,650] Liters
Company supermarket U [90,585; 100,650] Liters
Supermarket—A U [29,205; 32,450] Liters
Supermarket—B U [29,655; 32,950] Liters
Dtkp ;d = 2 Hypermarket—A U [76,905; 85,450] Liters
Hypermarket—B U [76,785; 88,650] Liters
Company supermarket U [99,585; 110,650] Liters
Supermarket—A U [32,805; 36,450] Liters
Supermarket—B U [32,355; 35,950] Liters
Rjk ;d = 1 Hypermarket—A 7.2 Km
Hypermarket—B 13.4 Km
Company supermarket 24.3 Km
Supermarket—A 20.1 Km
Supermarket—B 24 Km
Rjk ;d = 2 Hypermarket—A 23.9 Km
Hypermarket—B 26.4 Km
Company supermarket 4.2 Km
Supermarket—A 31.2 Km
Supermarket—B 28.1 Km
CT tjkp ;d = 1 Hypermarket—A 153 IDR/liter/km
Hypermarket—B 184 IDR/liter/km
Company supermarket 197 IDR/liter/km
Supermarket—A 137 IDR/liter/km
Supermarket—B 172 IDR/liter/km
CT tjkp ;d = 2 Hypermarket—A 131 IDR/liter/km
Hypermarket—B 144 IDR/liter/km
Company supermarket 148 IDR/liter/km
Supermarket—A 147 IDR/liter/km
Supermarket—B 146 IDR/liter/km
ET tjkp ;d = 1, 2 Hypermarket—A 0.0001 Lt.CO2-eq/lt./km
Hypermarket—B 0.0001 Lt.CO2-eq/lt./km
Company supermarket 0.0001 Lt.CO2-eq/lt./km
Supermarket—A 0.0001 Lt.CO2-eq/lt./km
Supermarket—B 0.0001 Lt.CO2-eq/lt./km

reached the optimal solution, namely, total costs in Indone- is the highest in the DCs at 64.56%, or 14,031 l, compared
sian Rupiah (IDR) of 18,036,770,000, total waste of 21,734 to the production factory warehouse. Next, Table 6 shows
L, and total environmental pollution of 109,526 ltrs.CO2-eq. the optimal solutions for the third objective function. The
Table 4 shows that the total cost of production and packag- highest pollutant is through refrigeration activities, with a
ing of milk carried out by the production factory is the larg- percentage of 93.2%, or 102,077 l ­CO2-eq of the total envi-
est cost component, namely, 46.9%, or IDR 8,475,626,161 ronmental emissions for transportation, cooling, and food
of the total cost. waste activities. Figure 3 shows that the total cost of activi-
Furthermore, the most significant cost component is the ties in the factory is 82.17% of the total supply chain costs
cost of cooling fresh cow’s milk products in the form of raw from suppliers to retail consumers, and the remainder is
materials purchased from suppliers before QC at the produc- costs incurred in the DC’s warehouse of 17.82% and total
tion factory, with a percentage of 21.6% IDR 3,905,525,270 environmental costs of 0.01%. It can be concluded that
of the total cost. Table 5 shows the optimal results for the the results of this research’s numerical calculations show
second objective function regarding food waste. It is known that the percentage of production costs for processing and
that the amount of milk that does not pass QC as food waste packaging of dairy products has the highest percentage of

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

Table 4  Optimal results on total No Component of the total cost Value (IDR) %
costs, carbon emissions, and
pollutant emissions 1 The milk purchasing cost at the production factory 1,083,206,337 6.01
2 The quality inspection cost from a supplier in the factory 60,561,155 0.34
3 The deterioration costs at the factory 509,594 0.00
4 The packaging costs at the factory 8,475,626,161 46.99
5 The cooling costs at the factory for the first inspection 3,905,525,270 21.65
6 The cooling costs at the factory for the second inspection 223,899,529 1.24
7 The holding cost of products at the factory 499,693,187 2.77
8 The distribution costs from the factory to DCs 570,941,571 3.17
9 The cost of milk quality inspection on DCs 208,082,031 1.15
10 The cost of waste on DCs 1,802,319 0.01
11 The cooling costs after inspection on DCs 559,343,654 3.10
12 The product holding costs on DCs 407,806,142 2.26
13 The transportation costs from DCs to consumers 2,037,488,340 11.30
14 The cost of transport emissions from factories and DCs 186,678 0.00
15 The cooling emission costs of factories and DCs 2,091,965 0.01
16 The waste emission costs at production plants and DCs 6,068 0.00
Total cost 18,036,770,000 100

Table 5  Optimal results on food waste such as water, sugar, powdered skim, and milk butter was
No Component of food waste Liters %
increased to 571,452 l. From the total, the quantity of
dairy products can be sold to supermarkets as follows: (a)
1 The amount of food waste in a pro- 7703 35.44 size 65-ml = 3,703,524 bottles; (b) size 125-ml = 962,916
duction factory pcs; (c) size 250-ml = 2,407,290 pcs; and (d) size 1000-
2 The amount of food waste in DCs 14,031 64.56 ml = 240,729 pcs. The optimal result of the decision vari-
Total food waste 21,734 100 able shown in Fig. 4 yields the total minimum cost, reduc-
ing food waste and environmental pollutants throughout
the echelon involved in the dairy supply chain.
the total cost, which is different from the results of Wangsa
et al.’s (2023) research. The total cost of packaging produc-
tion influences the highest cost; the result is the difference Sensitivity Analysis and Managerial Insight
between the previous studies. In this study, the parameter
for packaging production costs for 1 l was IDR 7583; com- Analysis of Centralized and Decentralized Model
pared to Wangsa et al.’s (2023) study, the cost of packaging Decisions
production was IDR 2083, or 27% lower than the unit cost
of production in this study. This subsection describes the evaluation of the results for
Figure 4 shows the optimal flow of a dairy milk sup- two scenarios, namely, centralized and decentralized deci-
ply chain from fresh raw milk to milk products. Start- sions. Table 7 shows the results of a comparison between
ing with the chain system, the total purchase of milk is single-objective and multi-objective functions. The study
256,043 l. Of that total, there is 7703 l that is in a rejected results for a single objective function have three scenarios,
state, so pure fresh milk that is only subsequently pro- namely, the single objective function of total costs, total
cessed is 248,340 l. Then, the production of dairy milk waste, and total environmental pollution. The results are
with some additional mixtures of additive compositions compared to the results of multi-objective functions. The

Table 6  Optimal results on No Component of environmental pollution Ltrs.CO2-eq %

Environmental pollution
1 The transport pollutant in factories and DCs 7,021 6.41
2 The pollutant of cooling in factories and DCs 102,077 93.20
3 The pollutant of food waste in factories and DCs 428 0.39
Total emission pollutant 109,526 100

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

Fig. 3  The component of the total cost

results of scenario 1 with minimizing the total cost show emission is in scenario 2, which is 114,549 l C
­ O2-eq, which
a result of IDR 18,036,770,000 with the same value when has a difference of 5,094 l C
­ O2-eq, or 4.65%, and 5,023 l
minimizing the multi-objective function is carried out. The ­CO2-eq, or 4.59%, compared to scenario 3 and the multi-
highest total cost in scenario 2 is when the mathematical objective function.
model has a single objective function to minimize total
food waste with a value of IDR 18,980,390,000, or the Analysis of Temperature Effect Setting
difference is IDR 943,620,000 (5.23%) compared to the on the Objective Function
multi-objective’s results.
The table shows that in the results of scenario 2, the The impact of temperature variance on total cost and waste is
objective function, namely, minimizing total waste, shows a shown in Fig. 5. The operator of the production factory per-
result of 16,517 l. Compared with the multi-objective func- forms his or her duties to set the cooling temperature for the
tion scenario results, we can find the optimal value with a storage of dairy products with a temperature range based on
difference of 5,217 l, or 24% lower than the multi-objective factory information as a reference temperature for cold stor-
optimization value of 21,734 l. Suppose a comparison is age of fresh milk, 4 to 7 °C, or 277 to 280 K, and the actual
made between the four scenarios. In that case, it is found temperature range of milk in the warehouse is 8 to 15 °C, or
that the highest total waste is in scenario 1, and the value 281 to 288 K. The figure shows that the total cost and total
is the same when multi-objective, with a value of 21,734 l. waste will increase if the temperature increases from 8 to
Furthermore, in scenario 3, minimizing total carbon 12 °C. The increase in total waste reached 2.32% between
emissions, the result is 109,455 l ­CO2-eq. This result of 71 l the values of 21,7334 l at 8 to 9 °C and 22,238 l at 10 to
­CO2-Eq. (0.065%) is lower than the multi-objective function 12 °C. The results show an increase in the total cost range of
of 109,526 l C­ O2-eq. In addition, the highest total carbon 7.8 to 60.6% with a value of IDR 18,036,770,000 and IDR

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

Fig. 4  The optimal value of solution

28,972,210,000 at 8 and 12 °C, respectively. From these The best choices are made based on key performance metrics
results, it was found that there was a difference in total costs for business sustainability, specifically the economic, social,
of IDR 1,408,230,000 (7.8%), IDR 3,318,640,000 (18.4%), and environmental aspects. Determining target functions,
IDR 6,183,780,000 (34.3%), and IDR 10,935,440,000 such as minimizing overall costs, reducing total food waste,
(60.6%). Based on the analysis, it can be concluded that and minimizing total carbon emissions, is a crucial require-
the increase in temperature is directly proportional to the ment for businessmen in order to preserve the sustainability
increase in total costs. The higher the actual temperature, the of the company. The study’s findings are useful in that they
higher the ratio between the performance coefficients, and show how optimizing production levels, inspection quality,
the higher the energy required to freeze, the higher the total and delivery processes can reduce total production costs,
supply chain costs obtained. This study’s results align with food waste, and carbon emissions (Table 4). Also, improv-
the results of Wangsa et al.’s (2023) research. ing the screening quality process can reduce food waste in
factories and distribution centers (Table 5). Then, the most
Managerial Insight significant reduction in carbon emissions comes from the
cooling and transportation processes (Table 6).
This study aims to assist professionals—in this case, produc- This study examines how temperature control affects total
tion managers—in making the best production decisions. costs and food waste. According to the study’s findings,

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

Table 7  Comparison of single- Objective function and decision variables Scenario 1–3 (single-objective) Scenario 4:
and multi-objective scenarios Multi-objec-
results 1 2 3 tive

Objective function
1. Total cost (IDR × 1,000,000) 18,036.77 18,980.39 18,933.54 18,036.77
2. Total waste (liters) 21,734 16,517 17,723 21,734
3. Total pollutant (ltrs.CO2-eq) 109,526 114,549 109,455 109,526
Decision variables
1. Milk purchased quantity (liters) 256,043 252,065 252,065 256,043
2. Quantity inspected at factory (liters) 248,340 248,340 248,340 248,340
3. Milk waste at the factory (liters) 7,703 3,725 3,725 7,703
4. Packaged milk (liters) 248,340 248,340 248,340 248,340
5. Milk products distributed to DCs (liters) 571,452 511,168 571,452 571,452
6. Quantity inspected at DCs (liters) 571,452 511,168 571,452 571,452
7. Milk waste at DCs (liters) 14,031 12,792 13,998 14,031
8. Milk sold to retailers (liters) 1,203,645 1,170,807 1,203,679 1,203,645
(a) Type of size 65-ml (bottles) 3,703,628 3,602,484 3,703,628 3,703,524
(b) Type of size 125-ml (pcs) 962,944 936,646 962,944 962,916
(c) Type of size 250-ml (pcs) 2,407,358 2,341,614 2,407,358 2,407,290
(d) Type of size 1-l (pcs) 240,736 234,162 240,736 240,729
1. Truck frequency from supplier to factory (trips) 21 21 21 21
2. Truck frequency from the factory to DCs (trips) 48 43 47 48
3. Truck frequency from DCs to retailers (trips) 100 100 100 100

Fig. 5  The impact of tempera- 25,000 28,972.21 30,000

ture on total cost and total waste
Total Waste 28,000
24,000 Total Cost
26,000

Total Cost (IDR x 10^6)

Total Waste (Liters)

23,000 24,000

22,000 22,000

20,000
21,000
18,000
18,036.77
20,000 16,000
8 9 10 11 12
Temp (oC)

increasing the temperature in storage will significantly the factory (backorder) and determining high-safety stock.
increase total costs and affect the amount of food waste. This This disruption will affect the company’s total costs. In tem-
research is in line with Wangsa et al. (2023). These findings perature regulation, several technologies can be used, includ-
can assist warehouse managers in maintaining product qual- ing intelligent technology in cold storage control systems
ity by controlling the digital thermostats that control storage (Guo 2020). The technology integrates with the Internet of
temperature. The production process will be disrupted if the Things (IoT), which can inform managers about real-time
digital thermostat is damaged or erroneous. Thus, several temperatures. So, this technology can minimize damage to
strategic decisions are needed, such as reordering milk from the product.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

Conclusions and Future Directions impact on the lack of milk supply to factories. Finally, social
aspects such as job opportunities can also be considered by
The MILP model in this study was successfully devel- increasing the number of dairy farmers.
oped in the Moreno-Camacho et al. (2023) model by add-
ing the total waste objective function and in the Wangsa
et al. (2023) model by adding a holding cost component
at the production plant and DCs. The results of numerical Appendix
computations with optimum global results for the multi-
objective model yield a total cost of IDR 18,036,770,000, Definitions of indices, parameters, and decision variables
a total waste of 21,734 l with food waste in the factory are as follows:
of 7703 l (35.44%) and DCs of 13,031 l (65.56%), and a Indices
total pollutant of 109,526 l ­CO2-eq with a total transport
pollutant of 7,021 l ­CO2-Eq. (6.41%), a total emission of i Suppliers, i 𝜖 I
refrigeration in warehouses of 102,077 l C ­ O2-Eq. (93.2%), j Distribution centers (DC), j 𝜖 J
and a total waste pollutant of 428 l ­CO2-eq. The variable k Retailers, k𝜖K
for the total quantity produced and packaged is 248,340 l; p Product type, p𝜖P
the total quantity purchased by production factories from t Time period, t𝜖T
milk suppliers is 256,043 l; the total quantity of food waste
at the factory is 7,703 l; the total quantity of food waste at Parameter
DCs is 14,031 l; and the total quantity sold by consumer
retailers is 1,203,645 l. Btjp Warehouse capacity for milk p at DCs j in
The findings indicate that the following sizes of dairy period t (liter)
products can be marketed to supermarkets such as 65 ml for CRtip Milk purchase cost p from suppliers i for milk
3,703,524 bottles; 125 ml for 962,916 pieces; 250 ml for p in period t (IDR/liter)
2,407,290 pieces; and 1 l for 240,729 pieces. The decision CStip Cost of inspection of milk p from supplier i in
variable’s optimal outcome produces the lowest cost while period t (IDR/liter)
minimizing food waste and environmental pollution across CPtp The cost of packaging of milk p in the produc-
the board for the entire dairy supply chain. The sensitiv- tion factory in period t (IDR/liter)
ity test is conducted to determine the impact of the storage CIRtip Cooling cost of milk p from supplier i at the
temperature of dairy products on total costs and waste. The production factory in period t (IDR /liter)
sensitivity test results with fresh milk storage reference tem- CHPtp Storage cost of milk p at the production factory
peratures of 8 to 12 °C showed an increase in total waste of in period t (IDR/liter)
2.32%. In addition, the temperature between 8 and 9 °C will CIPtp Cost of cooling milk p at distribution centers
give a total waste of 22,238 l at conditions of 10 to 12 °C, in period t (IDR/liter)
and the total cost will increase. Therefore, the total supply CDtjp Distribution costs from the production factory
chain cost and waste increase as the temperature increases. to DCs j for product p in period t (IDR/liter)
The study’s findings reveal that increasing storage temper- CKtjp Milk inspection costs p at DCs j in period t
atures can increase costs and food waste. It suggests that (IDR/liter)
warehouse managers control digital thermostats to maintain CHDCtjp Holding cost of milk p at DCs j in period t
product quality. Damaged or incorrect thermostats can dis- (IDR/liter)
rupt production, requiring strategic decisions such as back- CIDtjp The cost of cooling milk p from the production
ordering and high-safety stock levels. Intelligent technology factory to DCs j in period t (IDR/liter)
can help maintain storage temperatures. The temperature CTtjkp The cost of transporting milk p to DCs j to
control device must be in good condition so that there are no retailers k in period t (IDR/liter/km)
deviations from the measurement results. Therefore, control- CETtp Pollutant cost of transportation of milk p in
ling temperature is very important. period t (IDR/lt.CO2-eq)
Further study can be developed from the limitations of CEItp Inventory pollutant cost of milk p in period t
this study. First, the model considers the waste cost of dairy (IDR/lt.CO2-eq)
based on the distribution process of product decay from the CEDtp Food waste pollutant costs (dairy waste emis-
production process. Further research can involve inspec- sion costs) of milk p in period t (IDR/lt.CO2-eq)
tion errors at the production stage. Also, future studies can Dtkp Demand for milk production p to retailer k at
consider disruption from suppliers (farmers), which has an distribution centers j in period t (liters)

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

DRKUDtip The rate of decay of milk p from supplier i at ddctjp The amount of milk p distributed from the pro-
the production factor y in per iod t, duction factory to DCs j in period t (liters)
ekudtip The amount of dairy food waste of p from supplier
E
0 −( Y a )
DRKUDtip (%) = kpi ∙e tp

DRDCtjp The rate of decay of milk p from the production fac- i in the production factory in period t (liters)
tory at DCs j in period t, DRDCtjp (%) = kpj0 ∙ e−( ) edctjp The amount of milk waste p at DCs j in period t
Ea
Ztpj

ETtjkp Pollutant rate for the delivery of milk p to DCs (liters)

j and retailers k in period t (lt.CO2-eq) irtip The inventory of milk p for first inspection at the
EItp Pollutant rate for cooling milk p in period t (lt. production factory in period t (liters)
CO2-eq) iptp The inventory of milk p for second inspection at the
EDtp The rate of pollutants from milk waste p in production factory in period t (liters)
period t (lt.CO2-eq) idtjp The inventory of milk p at DCs j in period t (liters)
L Payload capacity on LTL trucks (liter) tettp Total pollutant of milk p during delivering in period t
T J K ��
PCRtp Milk production capacity p in the production (lt.CO2-eq), tettp =
∑ ∑∑ � �
ddctjp ⋅ Sj + qtjkp ⋅ Rjk ⋅ ET tjkp
t=1 j=1 k=1
factory in period t (liters) teitp Total pollutant of milk
�� p during cooling�in period
� t
Sj Distance from production factory to distribu- T I J
(lt.CO2-eq), teitp =
∑ ∑ ∑
irtip + iptp + idtjp ⋅ EI tijp
tion centers j (km) t=1 i=1 j=1

Rjk Distance from distribution centers j to retailers tedtp Total pollutant of milk
�� p from milk waste
� in period
� t
k (km)
T I J
(lt.CO2-eq), tedtp =
∑ ∑ ∑
ekudtip + edctjp ⋅ EDtijp
TELVt Threshold limit of pollutant in period t (lt. t=1 i=1 j=1

CO2-eq) ukudtip Quantity of milk p purchased by the production

Xtp Warehouse capacity for milk p at the produc- factory from supplier i in period t (liters)
tion factory in period t (liter) wdctjp The quantity of milk p inspected at DCs j in period t
Ycoldp The cold temperature of milk p in the produc- (liters)
tion plant in period t (°C) qtjkp The amount of milk sold p from DCs j to
Yhotp The hot temperature of milk p in the production retailer k in period t (liters)
factory in period t (°C) xxtp The quantity of milk p produced and packaged
Ytp The actual temperature of milk p in the produc- at the production factory for period t (liters)
tion plant in period t (°C) ylti The frequency of LTL trucks sent from supplier
Zcoldpj The cold temperature of milk p in distribution i to the production plant in period t (times)
centers j in period t (°C) ydtj The frequency of LTL trucks sent from the pro-
Zhotpj The hot temperature of milk p in distribution duction plant to DCs j in period t (times)
centers j in period t (°C) yctjk The frequency of LTL trucks sent from DCs j
Ztpj The actual temperature of milk p in distribution to retailers k in period t (times)
centers j in period t (°C) ytp Binary variable: ytp = 1 if production is done
at the production plant, ytp = 0 otherwise.
Decision variable zkudtip Quantity of milk p from supplier i inspected by
the production factory in period t (liters)
COPref1p The reference of performance coefficient for milk
p in the production plant, COPref 1p = Ycold p
Yhotp −Ycoldp

COPakt1p The actual performance coefficient for milk p Author Contribution Conceptualization: I.V., and N.A.J; methodology:
Y
in the production plant, COPakt1p = Yhot tp−Y I.V; investigation: I.V., I.D.W., and N.A.J; resources: N.A.J.; data cura-
p tp tion: I.V; and I.D.W; writing-original draft preparation: I.V., I.D.W.,
COPp COP ratio of milk p in the production plant, and N.A.J; writing-review & editing: I.V., and I.D.W; visualization:
COPakt1
COPp = COPref 1p N.A.J., and I.D.W.; supervision: I.V.; funding acquisition: NA; All
p authors read and approved the final manuscript.
COPref2pj The reference of performance coefficient for
Zcoldpj Data Availability All data supporting this study’s findings are available
milk p at DCs j, COPref 2pj = Zhot −Zcold from the corresponding author upon reasonable request.
pj pj

COPakt2pj The actual performance coefficient for milk p

Z
at DCs j, COPakt2pj = Zhot tpj−Z Declarations
pj tpj

COPdcjp COP ratio of milk p in DCs j, COPdcjp = COPakt2pj Competing Interests The authors declare no competing interests.
COPref 2pj

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Process Integration and Optimization for Sustainability

References International Dairy Federation (IDF) (2015) A common carbon foot-

print approach for the dairy sector. The IDF guide to standard life
cycle assessment methodology (Bulletin 479)
Bacchetti A, Bertazzi L, Zanardini M (2021) Optimizing the distribu-
Khazaeli S, Jabalameli MS, Sahebi H (2023) Bi-objective model
tion planning process in supply chains with distribution strategy
for multi-level supply chain by focusing on quality of agricul-
choice. J Oper Res Soc 72(7):1525–1538. https://d​ oi.o​ rg/1​ 0.1​ 080/​
tural products: a case study. Kybernetes. https://​doi.​org/​10.​
01605​682.​2020.​17277​85
1108/K-​05-​2022-​0745
BAPPENAS (2021) Food loss dan waste di Indonesia dalam rangka
Kompas (2022) Industri susu di Indonesia menurut ahli https://​www.​
mendukung penerapan ekonomi sirkular dan pembangunan ren-
kompas.​com/​food/​read/​2022/​09/​01/​18040​0275/​perke​mbang​an-​
dah karbon
indus​tri-​susu-​di-​indon​esia-​menur​ut-​ahli
Chaabane A, As’ad R, Geramianfar R, Bahroun Z (2022) Utilizing
Li P, Lan H, Saldanha-Da-Gama F (2019) A bi-objective capacitated
energy transition to drive sustainability in cold supply chains:
location-routing problem for multiple perishable commodi-
a case study in the frozen food industry. RAIRO-Oper Res
ties. IEEE Access 7:136729–136742. https://​doi.​org/​10.​1109/​
56(3):1119–1147. https://​doi.​org/​10.​1051/​ro/​20220​43
ACCESS.​2019.​29413​63
de Keizer M, Akkerman R, Grunow M, Bloemhof JM, Haijema R,
Mercier S, Villeneuve S, Mondor M, Uysal I (2017) Time–temperature
van der Vorst JG (2017) Logistics network design for perish-
management along the food cold chain: a review of recent devel-
able products with heterogeneous quality decay. Eur J Oper Res
opments. Comp Rev Food Sci Food Safe 16(4):647–667. https://​
262(2):535–549. https://​doi.​org/​10.​1016/j.​ejor.​2017.​03.​049
doi.​org/​10.​1111/​1541-​4337.​12269
Dorcheh FR, Rahbari M (2023) Greenhouse gas emissions optimi-
Mogale DG, Kumar SK, Márquez FPG, Tiwari MK (2017) Bulk wheat
zation for distribution and vehicle routing problem in a poultry
transportation and storage problem of public distribution system.
meat supply chain in two phases: a case study in Iran. Process
Comput Ind Eng 104:80–97. https://​doi.​org/​10.​1016/j.​cie.​2016.​12.​027
Integr Optim Sustain 7:1289–1317. https://​doi.​org/​10.​1007/​
Moreno-Camacho CA, Montoya-Torres JR, Jaegler A (2023) Sustain-
s41660-​023-​00339-6
able supply chain network design: a study of the Colombian dairy
Esteso A, Alemany MME, Ortiz Á (2021) Impact of product per-
sector. Ann Oper Res 324(1–2):573–599. https://​doi.​org/​10.​1007/​
ishability on agri-food supply chain design. Appl Math Model
s10479-​021-​04463-9
96:20–38. https://​doi.​org/​10.​1016/j.​apm.​2021.​02.​027
Mosallanezhad B, Hajiaghaei-Keshteli M, Triki C (2021) Shrimp
European Commission—Joint Research Centre (2015) Energy use in
closed-loop supply chain network design. Soft Comput 25:7399–
the eu food sector: state of play and opportunities for improve-
7422. https://​doi.​org/​10.​1007/​s00500-​021-​05698-1
ment; Publications Office: Luxembourg, ISBN 9789279482991
Nasr N, Akhavan Niaki ST, Seifbarghy M, Husseinzadeh Kashan A
Gharye Mirzaei M, Gholami S, Rahmani D (2023) A mathematical
(2022) An agri-fresh food supply chain network design with rout-
model for the optimization of agricultural supply chain under
ing optimization: a case study of ETKA Company. Adv Math
uncertain environmental and financial conditions: the case study
Financ Appl 7(1):187–198. https://d​ oi.o​ rg/1​ 0.2​ 2034/A ​ MFA.2​ 021.​
of fresh date fruit. Environ Dev Sustain 1–34. https://​doi.​org/​
19280​61.​1584
10.​1007/​s10668-​023-​03503-7
Naurah N (2022) Food Waste Index 2021: Indonesia Jadi Penghasil
Glover JL, Champion D, Daniels KJ, Dainty AJ (2014) An institu-
Sampah Makanan Terbesar Se-ASEAN," 25 Juli 2022. Available:
tional theory perspective on sustainable practices across the
https://g​ oodst​ ats.i​ d/a​ rticl​ e/f​ ood-w
​ aste-i​ ndex-2​ 021-i​ ndone​ sia-j​ adi-​
dairy supply chain. Int J Prod Econ 152:102–111. https://​doi.​
pengh​asil-​sampah-​makan​an-​terbe​sar-​se-​asean-​7FgZ2. Accessed
org/​10.​1016/j.​ijpe.​2013.​12.​027
25 Jul 2023
Goodarzian F, Kumar V, Ghasemi P (2022) Investigating a citrus
Patidar R, Agrawal S (2020) A mathematical model formulation to
fruit supply chain network considering CO2 emissions using
design a traditional Indian agri-fresh food supply chain: a case
metaheuristic algorithms. Ann Oper Res 1–57. https://​doi.​org/​
study problem. BIJ 27(8):2341–2363. https://​doi.​org/​10.​1108/​
10.​1007/​s10479-​022-​05005-7
BIJ-​01-​2020-​0013
Guo C (2020) Application of intelligent technology in cold storage
Purnomo MRA, Wangsa ID, Rizky N, Jauhari WA, Zahria I (2022) A
control system. In IOP Conference Series: Earth and Environ-
multi-echelon fish closed-loop supply chain network problem with
mental Science (Vol. 571, No. 1, p. 012026). IOP Publishing.
carbon emission and traceability. Expert Syst Appl 210:118416.
https://​i opsc​i ence.​i op.​o rg/​a rtic​l e/​1 0.​1 088/​1 755-​1 315/​5 71/1/​
https://​doi.​org/​10.​1016/j.​eswa.​2022.​118416
012026/​pdf. Accessed 12 Oct 2023
Rahbari M, Khamseh AA, Mohammadi M (2023a) A novel multi-
Handayani DI, Masudin I, Rusdiansyah A, Suharsono J (2021) Pro-
objective robust fuzzy stochastic programming model for sus-
duction-distribution model considering traceability and carbon
tainable agri-food supply chain: a case study from an emerging
emission: a case study of the Indonesian canned fish food industry.
economy. Environ Sci Pollut Res 30(25):67398–67442. https://​
Logistics 5(3):59. https://​doi.​org/​10.​3390/​logis​tics5​030059
doi.​org/​10.​1007/​s11356-​023-​26305-w
Hossain M, Jahan R (2019) Rice chain flow modeling for optimized
Rahbari M, Khamseh AA, Mohammadi M (2023b) Robust optimi-
supply and distribution network: the case of specialized rice mill
zation and strategic analysis for agri-food supply chain under
zones in Bangladesh. IJLSM 32(2):147–167. https://​doi.​org/​10.​
pandemic crisis: a case study from an emerging economy. Expert
1504/​IJLSM.​2019.​097582
Syst Appl 225:120081. https://​d oi.​o rg/​1 0.​1 016/j.​e swa.​2 023.​
Indrayani S (2021) Berita Pemerintahan: Industri Makanan dan
120081
Minuman Diakselerasi Menuju Transformasi Digital. Available:
Rajabi-Kafshgar A, Gholian-Jouybari F, Seyedi I, Hajiaghaei-Keshteli
https://​www.​komin​fo.​go.​id/​conte​nt/​detail/​33978/​indus​tri-​makan​
M (2023) Utilizing a hybrid metaheuristic approach to design an
an-​dan-​minum​an-​diaks​elera​si-​menuj​utran​sform​asi-​digit​al/0/​ber-
agricultural closed-loop supply chain network. Expert Syst Appl
ita. Accessed 10 Aug 2023
217:119504. https://​doi.​org/​10.​1016/j.​eswa.​2023.​119504

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Process Integration and Optimization for Sustainability

Validi S, Bhattacharya A, Byrne PJ (2014) A case analysis of a sustainable Yuniarti R, Masudin I, Rusdiansyah A, Handayani DI (2023) Model
food supply chain distribution system—a multi-objective approach. of multiperiod production distribution for closed-loop supply
Int J Prod Econ 152:71–87. https://​doi.​org/​10.​1111/​1541-​4337.​12269 chain considering carbon emission and traceability for agri-food
Vanzetti N, Broz D, Corsano G, Montagna JM (2018) An optimization products. Int J Indus Eng Oper Manag. https://​doi.​org/​10.​1108/​
approach for multiperiod production planning in a sawmill. Forest IJIEOM-​10-​2022-​0045
Policy Econ 97:1–8. https://​doi.​org/​10.​1016/j.​forpol.​2018.​09.​001
Wang CN, Nhieu NL, Chung YC, Pham HT (2021) Multi-objective Publisher's Note Springer Nature remains neutral with regard to
optimization models for sustainable perishable intermodal multi- jurisdictional claims in published maps and institutional affiliations.
product networks with delivery time window. Mathematics
9(4):379. https://​doi.​org/​10.​3390/​math9​040379 Springer Nature or its licensor (e.g. a society or other partner) holds
Wangsa ID, Vanany I, Siswanto N (2023) An optimization model for exclusive rights to this article under a publishing agreement with the
fresh-food electronic commerce supply chain with carbon emis- author(s) or other rightsholder(s); author self-archiving of the accepted
sions and food waste. J Ind Prod Eng 40(1):1–21. https://​doi.​org/​ manuscript version of this article is solely governed by the terms of
10.​1080/​21681​015.​2022.​20994​73 such publishing agreement and applicable law.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]