CHEMICAL ENGINEERING TRANSACTIONS
VOL. 76, 2019
A publication of
The Italian Association
of Chemical Engineering
Online at www.aidic.it/cet
Guest Editors: Petar S. Varbanov, Timothy G. Walmsley, Jiří J. Klemeš, Panos Seferlis
Copyright © 2019, AIDIC Servizi S.r.l.
ISBN 978-88-95608-73-0; ISSN 2283-9216
Integration of Analytic Network Process in Adaptive Lean and
Green Processing
Wei Dong Leonga, Bing Shen Howb, Sin Yong Tengc, Sue Lin Ngana, Wendy Pei
Qin Ngd, Chun Hsion Lime, Hon Loong Lama,*, Chee Pin Tanf, S.G. Ponnambalamg
aUniversity of Nottingham Malaysia Campus, Dept. of Chemical Engineering, 43500, Jalan Broga, Selemyih, Selangor Darul
Ehsan, Malaysia
bChemical Engineering Department, Faculty of Engineering, Computing and Science, Swinburne University of Technology
Jalan Simpang Tiga, 93350 Kuching, Sarawak, Malaysia.
cBrno University of Technology, Institute of Process Engineering & NETME Centre, Technicka 2896/2, 616 69 Brno, Czech
Republic
dCurtin University Malaysia, Dep. Chemical Engineering, CDT 250, 98009 Miri, Sarawak, Malaysia
eUniversiti Tunku Abdul Rahman, Sungai Long Campus, Jalan Sungai Long, Bandar Sungai Long, Cheras, 43000
Kajang,Selangor, Malaysia
fMonash University Malaysia, School of Engineering, Jalan Lagoon Selatan, 47500 Bandar Sunway, Selangor Darul Ehsan
Malaysia
gUniversity Malaysia Pahang, Faculty of Manufacturing Engineering, 26600 Pekan, Pahang Darul Makmur, Malaysia
HonLoong.Lam@nottingham.edu.my
The manufacturing and processing industry have been an important part of the global economy. Many industry
players are constantly looking for an alternative to improve their operation and environmental performance to
remain competitive in the market. The lean and green approach aims to reduce operation and environmental
waste within an organisation. In this study, a lean and green framework is proposed to evaluate the industrialist
performance to achieve higher performance efficiency and reduce environmental impact. Three main clusters
are incorporated in the framework such as environment, machine and resources. The analytic network process
(ANP) method is used to establish the relationship between the three clusters with the input from industry expert
from the respective field. A lean and green index is developed from the ANP model as a benchmarking for the
industrialist. Backpropagation method is utilized as the continuous analysis tools to analyse the performance of
the organization accordingly to the time step. The adaptive characteristic of backpropagation method is reflected
from the ability for continuous improvement with time. In this study, the lean and green index will be further
optimized with the adaptive approach. This paper proposes an adaptive model that can improve the industry’s
performance and practise continuous improvement through establishing the adaptive approach.
1. Introduction
Lean and green (L&G) approach has been a paradigm shift in the manufacturing sector. In the past decade, the
lean approach has shown many positive outcomes in the manufacturing industry. For example, the lean effect
reflected from Toyota practise is to produce the same vehicle with shorter time while maintaining the quality.
Leong et al. (2019b) mentioned that lean is very prevalent in big organisations such as Toyota, Boeing, Ford,
etc. The lean approach is defined as the reduction or elimination of non-value-added product from a production
process (i.e. Toyota) (Womack and Jones, 1994). On the other hand, the green approach is known as a strategy
that focuses on operation profitability by enforcing the proactive and environmentally-friendly process (Abdul-
Rashid et al., 2017). Having said that, the green approach minimizes or removes environmental waste from the
process.
The IEA (2018) reported that global energy demand has grown by 2.1 % in 2017, which was twice the rate of
2016. IEA (2018) also added that electricity and heat generation sector was the largest CO2 generator in the
industry sector. This requires effort from industrialists in improving their process to be L&G. As both L&G have
DOI: 10.3303/CET1976094
Paper Received: 02/04/2019; Revised: 19/06/2019; Accepted: 23/06/2019
Please cite this article as: Leong W.D., How B.S., Teng S.Y., Ngan S.L., Ng W.P.Q., Lim C.H., Lam H.L., Tan C.P., Ponnambalam S., 2019,
Integration of Analytic Network Process in Adaptive Lean and Green Processing, Chemical Engineering Transactions, 76, 559-564
DOI:10.3303/CET1976094
559
the same objectives in reducing and eliminating waste or non-value-added product from the operation, the
synergy between L&G can create greater impact simultaneously to the operation as well as the environment
(Melnyk et al., 2002). Leong et al. (2019a) added that both L&G approaches complement each other to enhance
the efficiency and effectiveness of the process compared with their individual applications.
In this paper, an adaptive method is proposed to improve the rate of L&G implementation. This study will focus
on developing a self-monitoring model for the industrialist to accelerate the implementation of L&G approach.
Analytic network process (ANP) is incorporated into the model to facilitate the industry expert input into the
model. The outcome of this model is expected to reflect the actual operating condition based through continuous
analysis of the operation and environmental waste reduction. The continuous analysis element reflects the
adaptation functionality of the model.
2. Methodology
2.1 Lean and green framework
The L&G framework, illustrated in Figure 1, acts as the systematic step to guide the industrialist towards L&G
processing. Leong et al. (2019a) mentioned that there are five main components to be considered for the L&G
framework such as manpower (MP), machine (MC), money (MY), material (MT) and environment (EV) (4M1E).
Figure 1: Lean and green framework for processing plants (Leong et al., 2019a)
The methodology of L&G approach:
a. From the interview with the plant manager, the critical parameters of 4M1E will be identified.
b. An operation goal will be established based on the expectation of the operation team.
c. Questionnaires will be distributed to the operation team for data collection.
d. Based on the collected data, an initial L&G index (LGI) will be developed as the benchmark for the
process.
e. The operation team will then set an LGI target to achieve for the next LGI analysis.
f. Based on the predetermined LGI target, the back-propagation optimisation model will analyse the
potential improvement for the process.
g. Data will be collected and analysed after improvement steps have been performed.
2.2 Lean and green index
In the processing sector, the characteristic of operation varies between different sector. Therefore, the priority
and importance of operation criteria are dependent on different processes. To get the priority and ranking of the
indicators for specific processing plants, the analytic network process (ANP) is used. Analytic hierarchy process
(AHP), introduced by Saaty (1980), is widely applied in research areas as well as business applications over
the past decades (How and Lam, 2017). ANP is the generic form of AHP that is used to analyse data that has
more complex, interdependent and feedback relationship among the relationship in the hierarchy (Ngan et al.
2018). The supermatrix approach is used in ANP to allow interaction between and within clusters in deriving the
final composite weightage for all components in the model (Saaty and Takizawa 1986). Ngan et al. (2018) further
560
added that the ANP approach assists decision-makers in determining the priorities of the indicator alternatives
by ranking the alternatives for specific goals.
Based on Figure 2, questionnaires will be distributed to a group of an industry expert in the field to provide their
insight and opinion. The input for the ANP-based questionnaire is based on a 9-point fundamental scale
illustrated in Table 1. The data will be obtained from the group of industrialists as the input to develop the
pairwise comparison matrix. Table 2 explains the indicators that are reflected in Figure 2. These indicators are
discussed and selected by the industry expert for this case study. The eigenvector generated from the
calculation will form the local priority matrix. Finally, a consistency ratio (CR) is calculated to ensure the
consistency of the input. According to Saaty (1980), CR of less than 0.10 is deemed to be acceptable.
The input of supermatrix is formed by combining the local priority matrix from ANP. The supermatrix reflects the
relationship between clusters and the elements in the model. Eq (1) shows the matrix of local priority for
individual cluster, while Eq (2) indicates the relationship between each cluster. If there is no relationship between
the cluster, the block matrix will be represented by 0. The weighted supermatrix is formed by normalizing the
unweighted supermatrix where is it transformed into a stochastic matrix where the sum of all columns is equal
to 1.
Figure 2: ANP model
Table 1: 9-point fundamental scale (Saaty, 2012)
Verbal Judgement Numeric Value
Extremely Important 9
Very Strongly more Important 7
Strongly more Important 5
Moderate more Important 3
Equally Important 1
Table 2: Description of indicators
No Indicator Description
1 MC-OEE Overall equipment effectiveness
2 EV-CO2 Carbon dioxide footprint
3 MY-OC Operation cost
4 MY-OP Operation profit
5 MP-OT Employee total overtime
6 MP-MC Medical leave
7 MP-KPI The achievable key performance index
8 MP-CR Employee competency rate
9 MP-LC Employee late check-in time
10 MP-ST Employee safety competency rate
11 MT-MI Total material/resource into the production
12 MT-PO Total product output
561
𝑚 =
𝑒1 𝑒2 ⋯ 𝑒𝑛
𝑒1 1 𝑤12 𝑤1… 𝑤1𝑛
𝑒2
⋮
𝑒𝑛
1/𝑤12
1/𝑤1…
1/𝑤1𝑛
1
1/𝑤2…
1/𝑤2𝑛
𝑤2…
1
1/𝑤𝑛…
𝑤2𝑛
𝑤…𝑛
1
(1)
𝑆 =
𝐿1 𝐿2 𝐿3
𝐿1 1 𝑆12 𝑆13
𝐿2
𝐿3
𝑆21
𝑆31
𝑆22
𝑆32
𝑆23
𝑆33
(2)
Based on the ANP outcome, the LGI is calculated as below:
𝐿𝐺𝐼 = 𝑤𝑀𝐶 × 𝑀𝐶 + 𝑤𝐸𝑉 × 𝐸𝑉 + 𝑤𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 × 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 (3)
where the 𝑤𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 , 𝑤𝑀𝐶 , and 𝑤𝐸𝑉 represent the weight of Resource, machine and environment respectively.
The weights are developed based on the cluster categorized in Figure 2.
MC is known as the heart of the processing or manufacturing as it contributes mainly by adding value to the raw
material to produce products. In the Machine (MC) cluster, MC focuses on the equipment that is used for
processing. One of the common specific components for MC is the overall equipment effectiveness (OEE). It
measures the availability, performance and product quality of the MC. The availability measures the actual
operation time of the process by monitoring setup and adjustment time, and equipment failure time. Next, the
performance of OEE mainly focuses on monitoring the idling and minor stoppage time and reduce speed time
of the equipment as this will slow down the production. The final factor of OEE is quality, which identifies the
total amount of defects being produced from the operation.
The second cluster indicates the environment (EV) factor as an important role in the manufacturing processing
industry. EV has a direct relationship contribution to global warming and climate change. Teng et al. (2019)
mentioned that global warming potential (GWP) could be a good indicator for EV and obtained state-of-art results
using novel statistical L&G optimization in a real refinery plant. Besides, Ng et al. (2014) have also shown that
environmental impact can be reduced with proper waste management. The common specific components used
in EV are carbon, water and solid emission. Many industrialists struggle to reduce waste emission and schedule
waste as the additional cost is required to handle the waste.
The last cluster is represented by resources which consist of manpower (MP), money (MY) and material (MT).
MP, MY and MT are categorized as resources as these are manipulating factors that can affect the performance
of the production. The MP indicator reflects the performance of human resources in the facilities. Indicators such
as key performance achievable (KPI), safety competency rate, medical leave, etc. are essential to evaluate the
condition and performance of the employee. It is adding on, MT targets on resources that are consumed within
the process such as raw material, by product and waste. In many cases, inventory management is critical to
processing facilities. The continuity of the operation highly depends on the availability of resources. This
indicator contributes to the sustainable development goal (SDG) 12, which focuses on responsible consumption
and production (SDG, 2019). Lastly, MY is one of the most critical components in across all component as it
reflects the feasibility and profitability of a processing facility. MY does not only reflect the profitability of the
facilities. It also indicates the Return On Investment (ROI) of the organization. Besides that, MY is also a
powerful factor that motivates the employee in improving their performance (Aguinis et al., 2013).
The priority of the indicators will be evaluated with the ANP model. The output from ANP will be further optimized
with the proposed adaptive model for continuous improvement.
3. Adaptive model
The initial LGI developed from the calculation will be the benchmarking point for the process. This will act as the
reference point for continuous process improvement. In many scenarios, static analytic methods are mainly
used in model prediction. The static model becomes ineffective and inaccurate when it copes with dynamic
conditions in the continuous processes.
The adaptive model constantly updates the data according to time and provide responsive action to the
industrialist. The adaptive model relies on the backpropagation (BP) algorithm to perform optimisation.
According to Griewank (2012), the BP algorithm is based on the reverse mode of differentiation. Rumelhart et
al. (1986) highlighted that the reverse mode of differentiation was popularized with an application such as a
neural network. Lerun (1998) has also shown the application of the backpropagation method in a neural network.
Gori and Maggini (1996) work have shown that BP method is able to achieve the local optimum solution.
562
With the L&G approach, during the improvement of the process, the expectation from L&G approach will
increase over time. There is a need to constantly monitoring and improve the process parameter in order to
meet the expected LGI outcome. The gradient descent optimisation model is used as the update rule. The error
between the expected LGI outcome and actual LGI outcome is calculated as below:
𝐸 =
1
2
(𝐿𝐺𝐼𝑎𝑐𝑐𝑒𝑠𝑠𝑒𝑑 − 𝐿𝐺𝐼𝑒𝑥𝑝𝑒𝑐𝑡 )
2
(4)
where E is the error, 𝐺𝐿𝐼𝑎𝑐𝑐𝑒𝑠𝑠𝑒𝑑 is actually obtained LGI while 𝐿𝐺𝐼𝑒𝑥𝑝𝑒𝑐𝑡 is the expected LGI. This will allow BP
to monitor and the LGI based on historical data. Based on the BP method, the weights of each cluster will be
back-calculated in Eq(5) while the indicators will be calculated in Eq(6).
𝑤𝑖+1 = 𝑤𝑖 −𝜂
𝐸𝑖 − 𝐸𝑖−1
𝑤𝑖 − 𝑤𝑖−1
(5)
𝑘𝑖+1 = 𝑘𝑖 −𝜂
𝐸𝑖 − 𝐸𝑖−1
𝑤𝑖 − 𝑤𝑖−1
×
𝑤𝑖 − 𝑤𝑖−1
𝑘𝑖 − 𝑘𝑖−1
(6)
where 𝑤𝑖 and 𝑘𝑖 are the weight of the current month, 𝑤𝑖+1 and 𝑘𝑖+1are the weightage for next month, 𝑤𝑖−1and
𝑘𝑖−1 are the weight of the previous month. 𝜂 is known as the learning rate where it is normally defined as 0.05
which also define as the changes of LGI will not be more than 5 %. The learning rate is critical in gradient
descent optimizer as larger learning rate will tend to cause the output to fluctuate. Larger fluctuation might
indicate that the learning rate is too large where the local optimum cannot be identified. Therefore, a smaller
learning rate should be used to obtain the optimum point.
Figure 3 demonstrates the BP method. Example, i1, i2 and i3 represents the 4M1E while o1 represents the LGI.
The expert inputs are generated from ANP and to be fixed in this method. The initial actual outcome is generated
as a baseline; the industry player can set an LGI target as the future achievable outcome. In the BP method, if
the targeted outcome is not equal to the actual outcome, the error percentage will be backpropagated to evaluate
the priority of the components to be improved in the next cycle. With the improvement, the weight of i1, i2 and
i3 can be improved to achieve better LGI. This process will repeat until the actual and target outcome is equal.
Figure 3: Illustration of backpropagation (BP) method Figure 4: L&G Index (Leong et al., 2019c)
Figure 4 illustrates the performance of L&G framework using AHP approach. The improvement in operating
performance can be observed through the LGI. The application of ANP method in L&G framework will
demonstrate the complexity and inter-relationship between all components. Thus, producing a more
comprehensive LGI representation of the organisation. Based on AHP approach, the deviation of LGI from BP
method is 1.3% between actual and targeted outcome.
4. Conclusions
The L&G approach in manufacturing processing has shown positive results from the industry feedback. The
need to strengthen the implementation effort of L&G from the industry is highly important. The development of
L&G framework can be a systematic guideline to the industrialist to assist them in implementing L&G practices.
The framework can also assist the industrialist in changing the operation behaviour from existing practices
towards L&G. The development of LGI does not only act as a benchmarking tool but also continuous
563
improvement tools for every industry sector. The backpropagation model will assist the industrialist by
recommending improvement steps by reflecting the important indicators by comparing to the expected and
targeted outcome. The gradient descent optimization model is used in backpropagation method where the
performance of LGI can be further improved. The future work on this research will be extended to review better
performing gradient descent method that can enhance the optimisation performance of the model. A case study
will be used to demonstrate the effectiveness of the model.
Acknowledgements
The authors would like to acknowledge financial support from the Ministry of Higher Education
(FRGS/1/2016/TK03/MUSM/01/1). Research funding and support from Newton Fund and the EPSRC/RCUK
(Grant Number: EP/PO18165/1) is also gratefully acknowledged. The authors also highly appreciate the
valuable input from the industry experts in providing industry experience and knowledge to this research paper.
References
Abdul-Rashid S.H., Sakundarini N., Raja Ghazilla R.A., Thurasamy R., 2017, The impact of sustainable
manufacturing practices on sustainability performance: empirical evidence from Malaysia, International
Journal of Operations & Production Management, 37 (2), 182-204.
Aguinis H., Joo H., Gottfredson R.K., 2013, What monetary rewards can and cannot do: How to show employees
the money, Business Horizons, 56(2), 241–249. DOI:10.1016/j.bushor.2012.11.007.
Gori, M., Maggini, M., 1996. Optimal convergence of on-line backpropagation. IEEE Trans Neural Networks.
DOI: 10.1109/72.478415
Griewank A., 2012, Who Invented the Reverse Mode of Differentiation? Documenta Mathematica, Extra Volume
ISMP, 389-400.
How B.S., Lam H.L., 2017, Integrated Palm Biomass Supply Chain toward Sustainable Management. Chemical
Product and Process Modeling, 12(4) DOI: doi.org/10.1515/cppm-2017-0024.
IEA 2018, CO2 Emissions. , accessed 28.12.2018.
Leong W.D., Lam H.L., Lim C.H., Tan C.P., Ponnambalam S.G., 2019a, Outcome-based approach to lean and
green manufacturing processing, Chemical Engineering Transactions, 72, 361-366
DOI:10.3303/CET1972061.
Leong W.D., Lam H.L., Ng W.P.Q., Lim C.H., Tan C.P., Ponnambalam S.G., 2019b, Lean and Green
Manufacturing – A Review on Application and Impact, Process Integration and Optimisation for
Sustainability, 3(1), 5-23.
Leong W.D., Teng S.Y., How B.S., Ngan S.L., Lam H L., Tan C.P., Ponnambalam S.G., 2019c, Adaptive
Analytical Approach to Lean and Green Operations. Journal of Cleaner Production.
DOI:10.1016/j.jclepro.2019.06.143
Melnyk S., Sroufe R., Calantone R., 2002, Assessing the effectiveness of US voluntary environmental
programmes: An empirical study, International Journal of Production Research, 40(8), 1853-1878.
Ng W.P.Q., Lam H.L., Varbanov P.S., Klemeš J.J., 2014, Waste-to-Energy (WTE) network synthesis for
Municipal Solid Waste (MSW), Energy Conversion and Management, 85, 866–874. DOI:
10.1016/j.enconman.2014.01.004.
Ngan S.L., Promentilla M.A.B., Yatim P., Lam H.L., 2018, A Novel Risk Assessment Model for Green Finance:
the Case of Malaysian Oil Palm Biomass Industry. Process Integration and Optimization for Sustainability,
3(1), 75-88. DOI:10.1007/s41660-018-0043-4.
Rumelhart D.E., Hinton G.E., Williams R.J., 1986, Learning representations by back-propagating errors. n
Nature, 323, 533-536.
Saaty T.L., 1980, The Analytic Hierarchy Process, McGraw-Hill, New York, United States.
Saaty T.L., 2001, The analytic network process. RWS Publications, Pittsburgh, United States.
Saaty T.L., Takizawa M., 1986 Dependence and independence from linear hierarchy to nonlinear network. Eur
J Oper Res, 26, 228–237.
SDG, 2019, Sustainable Development Goals. Sustainable Development Knowledge Platform,
, accessed 02.11.2019.
Teng S.Y., How B.S., Leong W.D., Teoh J.H., Siang Cheah A.C., Motavasel Z., Lam H.L., 2019. Principal
component analysis-aided statistical process optimisation (PASPO) for process improvement in industrial
refineries. Journal of Cleaner Production, 225, 359-375.
Womack J., Jones D.T., 1994, From lean production to the lean enterprise, Harvard Business Review, 72, 93-
104.
564