Assessing Drivers Influencing Net-Zero Emission Adoption in Manufacturing Supply Chain: A Hybrid ANN-Fuzzy ISM Approach

Abstract

Nowadays, there is a constant focus on implementing the net-zero emission (NZE) concept in the manufacturing supply chain (MSC). To reduce emissions and improve organisational efficiency, adopting the net-zero concept is a prevalent trend in today’s highly competitive global business environment. Governments and stakeholders are pressuring the manufacturing sector to use natural resources efficiently and reduce environmental impacts. As a result, the manufacturing industry is focusing on cleaner production using net-zero practices. This study aims to identify and analyse the interaction among the drivers of net-zero adoption in the MSC. Through a systematic literature review (SLR), a list of drivers was recognised. To validate these drivers, we conducted an empirical study with 173 respondents from the Indian manufacturing industry. Further, we employed an artificial neural network (ANN) to weigh the nonlinear effect of drivers. Fuzzy interpretive structural modelling (F-ISM) was used to identify the interaction relationships among the drivers and construct a hierarchical structure among these identified drivers. The fuzzy matrix of cross-impact multiplications applied to the classification (F-MICMAC) method was used to categorise these drivers into driving and dependent categories. The outcomes of ANN show that Environmental predictors (100%) emerged as the most significant drivers, followed by Economic drivers (60.38%) and Technological drivers (59.05%). This study is a valuable resource for academia and industry professionals, providing essential insights into how adopting net zero facilitates the manufacturing industry’s ability to achieve net zero across the supply chain.

1. Introduction

In recent years, environmental sustainability has emerged as a pivotal strategic objective within manufacturing supply chains (MSCs) globally [1]. Rapid industrialisation and consequent economic growth have brought significant ecological challenges, including heightened greenhouse gas (GHG) emissions, climate change impacts and depletion of the ozone layer [2]. These challenges underscore the urgent need for proactive measures among business professionals towards environmental stewardship [3]. Manufacturers are increasingly compelled to balance economic interests across stakeholders while enhancing ecological efficiency and fostering a competitive business environment [4]. The ‘United Nations Climate Change Conference’ (COP-26) has underscored the imperative of sustainable development, advocating a shift from economically driven manufacturing to low-emission production methods to achieve a net-zero society [5]. This global agenda necessitates integrated efforts within organisations, spanning strategic, tactical and operational levels of SC management, to achieve environmental sustainability [6,7]. In response to these challenges, frameworks such as net-zero emission (NZE) regulatory policies are crucial to curb emissions and mitigate environmental impacts. Aligned with the United Nations Sustainable Development Goal 2030 agenda, NZE is designed to balance environmental protection, economic growth and social welfare [8]. NZE plays a pivotal role in guiding MSCs towards sustainable development by imposing regulations that promote emission-efficient operations. Economies worldwide have committed to achieving NZE targets, compelling industry leaders and decision makers to effectively manage supply chain (SC) operations [9]. Key strategies include reducing production costs, fostering international collaborations, enhancing product quality and attracting foreign investments [10]. Countries like China, the USA and India, among others, have set ambitious targets for achieving NZE, requiring robust management of SC operations and business activities [11]. These targets necessitate a comprehensive NZE that targets significant stakeholders such as suppliers, logistics providers, manufacturers and retailers. These policies encourage greener practices and innovative resource conservation within manufacturing processes. Governments worldwide have introduced market-based regulatory mechanisms like pollution taxes and cap-and-trade policies to incentivise organisations towards sustainable operations [12]. Such policies include carbon taxes, carbon offset programs and carbon cap-and-trade systems aimed at enhancing the carbon efficiency of manufacturing operations [13].

In India, the implementation of new policies and sustainable initiatives in the manufacturing sector is closely monitored by regulatory bodies, reflecting a growing awareness and responsiveness to environmental concerns among stakeholders [14]. Despite these advancements, challenges persist, including economic uncertainties and technical complexities associated with adopting NZE [15]. Successful implementation requires collaboration and support from suppliers, stakeholders, consumers and governmental bodies [16]. The Indian manufacturing sector, poised as one of the fastest-growing globally, holds substantial revenue generation and job creation potential. However, realising this potential hinges on the sector’s ability to effectively align with environmental regulatory policies [17]. Large enterprises, leveraging financial resources, advanced technology and robust infrastructure, are at the forefront of adopting sustainable practices.

Despite their significant socio-economic contributions, the manufacturing sector requires enhanced government support to navigate towards an environmental sustainability roadmap [18]. Identifying key drivers influencing their implementation is critical to fostering broader adoption of NZE among the manufacturing sectors. Such insights can empower manufacturing organisations, particularly in emerging economies, to strategize their environmental initiatives effectively and advance towards long-term sustainability goals. Despite the growing recognition of NZE’s importance, there remains a dearth of literature on its adoption within emerging economy manufacturing sectors [14].

Addressing these gaps necessitates the development of analytical frameworks that elucidate the critical drivers influencing NZE implementation in manufacturing SCs. Such frameworks are essential for analysing the interrelationships among these drivers and devising targeted strategies to facilitate the widespread adoption of NZE. In this regard, adopting net zero to promote sustainability practices within the MSC will benefit the manufacturing sector. Identifying and analysing the interaction among drivers that can support the net-zero business model will improve market competitiveness and the overall effectiveness of the MSC. Therefore, the present study aims to develop a framework for modelling the associations within the potential drivers identified by a systematic literature review (SLR). Further, the present study will influence the decision to adopt net zero in the MSC. Based on the above premises, the following research questions (RQs) are proposed:

RQ1: What are the key drivers that facilitate the adoption of NZE in the MSC?

RQ2: What nonlinear relationships exist among these identified drivers, influencing the adoption of NZE?

RQ3: How do these drivers interrelate in influencing net-zero adoption?

RQ4: What are the driving and dependence powers of these drivers in the context of net-zero adoption?

Based on the SLR, an expert panel identified and validated a list of potential drivers. Using the survey data, an artificial neural network (ANN) was employed to estimate the nonlinear relationship among drivers. The fuzzy ISM method is used to interrelate the drivers, and the fuzzy MICMAC method was used to categorise the drivers of net-zero adoption into different categories.

The upcoming sections of the present study are organised as follows: Background and related literature are presented in Section 2. An overview of the research methodology is presented in Section 3. Section 4 outlines the development of the framework. In Section 5, we discuss the data analysis and results. Section 6 represents a discussion of the results. Lastly, Section 7 explores the conclusions, limitations and scope of future research.

2. Background and Related Literature

Extracting useful information from numerous scientific articles available in various databases and inaccessible to practitioners and policymakers is a challenging and time-consuming task that requires significant resources [19]. Therefore, it is crucial to provide policymakers, practitioners and academics with a detailed overview of the manufacturing SC and NZE to guide future research. The strategy used to identify the drivers from the SLR is represented in Figure 1.

To obtain article-based data, three indexed databases, namely Scopus, Web of Science and IEEE Xplore, were used. These selected databases are widely used for quantitative analysis [19]. The flow chart depicted in Figure 2 shows the main keywords used for article selection. Further, Figure 3 shows the various inclusive and exclusive criteria considered for article selection.

2.1. Theoretical Background

The increase in global temperatures due to GHG emissions has become a critical issue requiring definitive solutions. In 2020, the Earth’s surface temperature was 0.98 °C above the 20th-century average [7]. China is recognised as the most significant GHG emitter, followed by India and the United States. Industrially developed countries have seen a rise in GHG emissions, highlighting the need for sustainable and green organisational processes to achieve NZE’s aim through rapid and radical structural change. Achieving the NZE goal is expected to significantly impact the climate change landscape. However, NZE’s goal will require overcoming significant challenges, particularly in managing global SC for large organisations [3,6].

The core principles of various frameworks remain the same to attain sustainable practices that individuals and large-scale industries can adopt globally. Decarbonising the SC is crucial, as GHG emissions have increased significantly in upstream SCs. Direct operational emissions may be low in many sectors, but supplier emissions can be up to 10 times higher. Therefore, managing complexity in the global SC is essential to achieving NZE goals and establishing a net-zero SC management framework. Sustainability has become increasingly important for firms seeking a competitive advantage alongside societal growth. Information management is viewed as a new development paradigm aiming to improve compliance with social, environmental and economic sustainability regulations. A net-zero SC framework can manage the GHG emission surge following industrialisation. The core principles of such a framework involve identifying and eliminating strategies to integrate environmentally sustainable practices. These practices encompass ‘material sourcing’, ‘product design’, ‘selection processes’, ‘manufacturing’ and ‘delivery’.

2.2. Net-Zero Adoption and MSC

The concept of a net-zero supply chain has garnered significant attention in recent years as organisations recognise the urgent need to mitigate climate change [20]. According to the United Nations, net zero entails ‘cutting GHG emissions to as close to zero as possible, with any remaining emissions re-absorbed from the atmosphere, by oceans and forests, for instance’ [5]. Industrial logistic networks significantly affect emissions across various supply chain stages, including raw material extraction, manufacturing, transportation and end-of-life disposal. The global supply chain substantially contributes to the world’s GHG emissions. Although pinpointing an exact percentage is challenging, studies suggest that emissions from global supply chains are substantial [21,22]. A report by Accenture states that global supply chains are responsible for up to 60% of the world’s GHG emissions [11].

Global institutions are collaborating to create a more net-zero emission environment. The manufacturing sector is moving towards a NZE economy [23,24]. Achieving NZE in the MSC involves minimising or offsetting the GHG emissions associated with the entire supply chain network, including the production, transportation and disposal of goods. This concept requires measuring, reducing and offsetting GHG emissions throughout the supply chain network [25,26]. Numerous studies have explored strategies for reducing GHG emissions in supply chains, as represented in Figure 4.

To adopt and implement these strategies effectively, collaboration among all stakeholders, including governments, businesses and consumers, is crucial for driving sustainable practices and reducing GHG emissions in global supply chains. A promising approach to achieving these goals is the implementation of digital technology, which can facilitate the transition to a net-zero economy.

2.3. Literature Gaps

In recent years, the adoption of the net-zero concept in the manufacturing industries has gained attention due to its crucial role in promoting an emission-free supply chain and reducing waste, resource extraction and pollution. These practices incorporate a zero-waste philosophy and consider closed-loop supply chain flows within the system. Studies have shown the benefits of adopting the net-zero concept, including improved organisational performance, enhanced market competitiveness and better resource utilisation. Compared to traditional concepts like reusing, reducing and recycling, the manufacturing sector’s approach now extends to new strategies such as regeneration and restoration. Despite these benefits, there remains a need for a better understanding of the net-zero concept in the manufacturing sector, particularly regarding its role and benefits for the industry in developing economies. Due to government pressure and dynamic customer markets, the manufacturing sector in emerging economies recognises the need for net-zero techniques to improve ‘resource efficiency’, ‘decentralised decision making’ and ‘energy management’. These technologies also foster collaborative environments within SC networks to facilitate the net-zero goal. Therefore, understanding the role of these net-zero adoption drivers is crucial for developing economies. We identified key drivers that can support an MSC, as shown in Figure 5. Based on an SLR, gaps related to the relationship between net-zero adoption and the MSC were identified:

• Existing studies provide a theoretical understanding of the NZE concept and MSC, but empirical investigations are needed in developing countries, necessitating further research.
• While the studies discuss net-zero economy enablers, the impact on improving organisational performance in the manufacturing industry remains to be determined.
• Current research overlooks the manufacturing perspective in emerging economies. Given the varied levels of adoption and maturity among MSCs, a detailed understanding of these concepts is needed.

2.4. Motivation and Novelty

Addressing climate change necessitates adopting net zero in MSCs, yet existing research largely overlooks the complex interactions among drivers influencing this transition. Traditional studies focus on individual drivers, using primary methodologies that fail to capture the nuanced interplay between factors. This research fills this gap by introducing a novel hybrid approach that integrates ANN, fuzzy ISM and fuzzy MICMAC analyses. ANN excels at modelling complex, nonlinear relationships among multiple drivers, providing a sophisticated analysis of their interactions. Fuzzy ISM is adept at handling the inherent uncertainty and ambiguity in expert opinions, ensuring that the model accurately reflects the complexity of real-world scenarios. MICMAC analysis categorises drivers based on their ‘driving’ and ‘dependence power’, offering strategic insights that facilitate more effective decision making. The novelty of this research lies in its unprecedented hybrid approach, combining ANN, fuzzy ISM and MICMAC for the first time in the context of net-zero adoption in MSCs. This comprehensive and robust analytical framework surpasses the limitations of previous studies by providing a holistic understanding of the intricate interactions among drivers. By leveraging the strengths of each methodology, the study not only enhances our experience but also offers a scalable and adaptable model for future research. This hybrid model sets a new standard in the field, bridging critical research gaps and offering valuable insights for policymakers and industry stakeholders aiming to achieve net-zero targets through more effective and sustainable practices.

3. Methodology

We developed the proposed research model by employing a combination of quantitative and qualitative research methods for data collection and a comprehensive three-step ANN-fuzzy ISM-MICMAC method for data analysis, as illustrated in Figure 6 and detailed in the subsequent sections. The first phase involved conducting an SLR to identify potential drivers of net-zero adoption in the MSC. In the second phase, we used a hybrid survey approach, incorporating both online and offline questionnaires, to gather data for analysing nonlinear relationships through ANN. The third phase involved an expert panel review to examine the structural associations among the identified drivers. Finally, we utilised MICMAC analysis to categorise these drivers based on their driving and dependence power. We offer strategic insights for more effective decision making and implementing net-zero initiatives in the MSC.

The drivers identified from the literature were validated by a survey conducted within the Indian manufacturing industries, ensuring that the selected drivers significantly influence the MSC in emerging economies.

A two-step multi-investigative approach was applied using the ANN-fuzzy ISM technique. F-ISM, which only checks linear associations, is unsuitable for decisions like technology adoption. Therefore, ANN was integrated with F-ISM to detect linear and nonlinear relationships among drivers for adopting the net-zero technique in MSC. F-ISM identifies linear relationships where one driver influences another. At the same time, ANN explores nonlinear relationships and learns and identifies complex interactions.

3.1. Empirical Analysis

An empirical analysis using quantitative and qualitative methods offers a solid theoretical foundation for the study. Additionally, exploratory factor analysis (EFA) was employed to organise the drivers and analyse the questionnaire survey data collected from respondents.

3.2. Details of the Survey

To gather empirical data for our study, a comprehensive survey was designed and administered to participants from the manufacturing sector. The following subsections provide a detailed description of the survey design, distribution and data collection process, including the specific methodologies used to ensure the reliability and validity of the collected data.

3.2.1. Questionnaire Survey Development and Data Gathering

In the first phase, an empirical investigation was conducted to establish statistical validity and provide a theoretical foundation for the drivers identified from the SLR. Following this, a five-point Likert scale (1: No Influence (NI); 5: Max Influence (MI)) questionnaire survey was developed based on prior research [27]. The questionnaire underwent pre-testing by three professors: one from the management department and two from the industrial engineering department and four experts from manufacturing industries specialising in NZE and supply chain consultancy (see

Table 1. Demographics of experts (Expt.) participating in questionnaire pre-testing.

Expert	Background	Work Experience (in Years)
Expt1	Academia	35
Expt2	Academia	28
Expt3	Academia	19
Expt4	Manufacturing head	23
Expt5	General manager	20
Expt6	Production manager	15
Expt7	Logistics head	16

). After detailed discussions with these experts, several modifications were made based on their feedback. The sampling process and eligibility criteria for selecting organisations for this study are detailed as follows:

3.2.2. Sampling and Eligibility Criteria

To select manufacturing organisations for the study, we employed a systematic sampling approach comprising the following steps:

Sampling Frame: We identified a list of manufacturing industries from industry databases and government records, which served as the foundation for selecting potential study participants.

Inclusion Criteria: Industries had to meet criteria including size, the presence of an operational website, the nature of operations and the adoption of net-zero practices. We targeted small and medium-sized manufacturing industries that had either implemented or were in the process of implementing net-zero sustainability practices, ensuring relevance to the research objectives.

Exclusion Criteria: Industries that failed to meet the inclusion criteria, were unwilling to participate or provided incomplete or unreliable data were excluded to ensure data quality.

The questionnaire consisted of three phases: (1) Challenges, (2) Driving Forces and (3) Maturity Items. The same population was surveyed across all three phases. This study reports only the results related to the drivers. We prepared a list of 732 manufacturing firms from various geographical locations in India using industry directories. These firms were contacted via mail, which explained the survey’s purpose and provided a description of the drivers involved. Initially, only 37 responses were received within the first month. We then leveraged LinkedIn to reach out to top management employees from the listed firms, urging them to participate. Ultimately, we finalised 173 responses after eliminating 9 biased ones. With a response rate of 23.7%, this outcome is considered acceptable and comparable to other empirical investigations in the Indian context [5,27]. Responses were kept anonymous to minimise bias and ensure the integrity of the primary data collected. A brief description of the drivers was included in the email to enhance understanding and clarity. The details of the respondents are shown in

Table 2. Respondents’ demographic summary.

Demographic Characteristics		Subcategory with Frequency and Percentage
Qualification	Bachelor’s	Master’s	PhD
	93	55	25
	53.7%	31.8%	14.5%
Work experience (years)	Junior (1 to 5)	Intermediate (6 to 10)	Senior (Above 10)
	64	82	27
	37%	47.4%	15.6%
Background	Manager	Supervisor	Researcher
	77	33	63
	44.5%	19%	36.5%
Respondent organisation	Automobile, rubber and textile industries	Agriculture and food processing	Others
	92	37	44
	53.2%	21.4%	25.4%
Industry type	Small	Medium	Large
	54	85	34
	31.2%	49.2%	19.6%

.

3.2.3. Reliability and Validity Checks

SPSS v26.0 software was used to conduct reliability and validity tests, ensuring the accuracy and integrity of the data gathered from the manufacturing sector. Reliability was assessed using Cronbach’s alpha, which yielded a value of 0.863. This is considered acceptable, according to the literature [28]. Convergent validity was determined by examining the factor loadings of each variable, with values exceeding 0.5 indicating validity [29]. In this study, all factors exhibited factor loadings greater than 0.5, confirming the gathered data’s convergent validity.

3.3. Exploratory Factor Analysis

The EFA approach is frequently employed in operations management to determine the structure of factors, especially when there is limited or no existing literature or specific concepts available. EFA offers several advantages, such as minimising information loss and reducing significant variables into a more manageable structure. To assess the suitability of the data for EFA, we conducted Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin (KMO) measure. Bartlett’s test of sphericity should have a p-value less than 0.01 and the KMO value should be at least 0.60. In our study, the KMO value was found to be 0.841, indicating that the data were appropriate for EFA. Using the varimax factor rotation method, the EFA revealed that the drivers of NZE in the MSC could be categorised into eight major groups, accounting for 67.725% of the total variance. The factor loadings for these drivers ranged from 0.603 to 0.906, exceeding the acceptable threshold suggested in the literature.

4. Development of a Framework for MSCs

The EFA analysis indicated that all drivers related to NZE are highly significant for this study. The expert team reviewed the EFA results and was then asked to categorise these drivers into groups for framework development. Based on the inputs from experts, we named these groups as follows: (1) Regulatory and Policy, (2) Economic, (3) Technological, (4) Market and Consumer, (5) Supply Chain and Procurement, (6) Corporate Governance and Ethics, (7) Environmental and (8) Operational. Figure 7 illustrates the detailed framework developed for the MSC after the EFA analysis. To understand expert perceptions of the key drivers for net-zero adoption, these drivers were analysed using ANN, F-ISM and MICMAC techniques through an expert panel review. Data were collected from three types of experts: industry specialists, academic experts and those experienced in both areas. Experts assessed each pair of factors for interrelationships, identifying one of four possible connections: ‘achieved by’, ‘leads to’, ‘bidirectional’ or ‘no relation’. These assessments were then mapped following the guidelines in Table 1.

To provide a clear understanding, we categorised the identified drivers into eight main groups, which are predictors in the context of ANN and their respective subgroups. Using the collected data as a foundation, ANN was employed to assess the nonlinear influence of these predictors. The scale used in this study for measuring various drivers and predictors. All drivers and predictors were evaluated using 173 and 91 responses through a questionnaire survey.

4.1. Case Study

A case study was performed on the developed framework within a leading manufacturing industry in India. This section provides a comprehensive discussion of the details of the case study.

Details of Case Organisation

A rubber tire manufacturing industry (disguised as ABC) was selected to test the framework due to its critical role in supporting net-zero practices. The case industry, which was founded in 1983 in northwestern India, employs approximately 500 people. ABC has implemented green business practices, manufacturing and supplying rubber goods to large corporations and international clients. The company prioritises sustainability initiatives and regularly conducts employee training and awareness programs. After discussions with top management, ABC agreed to participate in the study. An expert team of six members was formed, including one senior manager, one R&D manager, one logistics head, one plant head, one operations manager and one manufacturing manager. Each team member has over 10 years of experience in the MSC, and three are actively involved in sustainability projects.

4.2. Artificial Neural Network

In the first step of ANN-F-ISM modelling, ANN is chosen for its ability to handle input data without requiring a specific equation form. It adapts easily to different datasets and effectively manages incomplete or partial data [30]. ANN offers higher predictive precision than traditional linear models like structural equation modelling (SEM), multiple linear regression, multiple discriminant analysis (MDA) and binary linear regression analysis. F-ISM is a valuable method for predictors with a statistically significant effect on dependent variables (net-zero adoption). However, traditional linear statistical techniques, including sequential ISM, can only detect linear relationships, oversimplifying complex human decision making [31,32,33]. To address this, we used ANN, a leading artificial intelligence (AI) model, in the decision-making process. ANN, especially ‘the multi-layer perceptron type’, can model ‘nonlinear relationships’ and learn from input/output mappings [34]. ANN models mimic the structure of the human brain, detecting nonlinear and non-compensatory relationships [35,36]. Compared to linear models, ANN achieves superior predictive accuracy, adaptability and robustness [34,37,38]. However, we developed a two-step method combining ANN with fuzzy ISM since ANN is unsuited for examining causal relationships and hypothesis testing [9,39,40].

4.2.1. Training of ANN

The ANN training phase aims to adjust the network’s internal weights to accurately capture the implicit relationship between inputs and outputs. The provided input/output pairs are formatted as follows:

T = (a₁, x₁), (a₂, x₂), .., (a_ni, x_ni)

(1)

where x_i indicates the ith sample of the input parameter vector and a_i denotes the corresponding vector of output responses. These samples implicitly define the nonlinear relationship between the input parameters and output responses. The objective is to develop an ANN model to learn this implicit function. The general form of the ANN output is expressed as:

u = u(x, w)

(2)

where ‘w’ is the vector of unknown weights of the ANN, ‘x’ is the input parameters and ‘u’ is the vector of output responses of the ANN. The vector of unknown weights is determined by solving an optimisation problem.

$$\mathrm{w}\mathrm{*}=\min \mathrm{x} \mathrm{E}\mathrm{t}=\min \mathrm{x}{\sum }_{\mathrm{i}}\parallel {\mathrm{a}}_{\mathrm{i}}-\mathrm{y}({\mathrm{x}}_{\mathrm{i}}-\mathrm{w})\parallel$$

(3)

where Et represents the error across all samples. Many approaches can be used to solve this optimisation problem.

$${\mathrm{w}}_{\mathrm{n}\mathrm{e}\mathrm{x}\mathrm{t}}={\mathrm{w}}_{\mathrm{n}\mathrm{o}\mathrm{w}}-\mathsf{\beta }{\displaystyle \frac{\mathsf{\mu }\mathrm{E}\mathsf{\tau }}{{\mathsf{\mu }}_{\mathrm{w}}}}$$

(4)

where β is referred to as the ‘ANN’ learning rate [32]. The weights are initially set to random values. This process continues until a solution to Equation (3) is found.

Continuous optimisation and calibration of ‘w_i’ and ‘c_i’ are essential to achieve accurate function training. The goal is to minimise the mean square error to its optimum. This iterative process repeats until the required level of accuracy is attained. According to Raut et al. (2018) [36], the calibration procedure for weights ‘w_i’ and biases ‘c_i’ is as follows:

$${\mathrm{V}}_{\mathrm{i}}={\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{u}}_{\mathrm{i}}+{\mathrm{c}}_{\mathrm{i}}$$

(5)

The bias ‘c_i’ represents a nonzero value added to the sum of inputs multiplied by their corresponding weights. In this sum, ‘V_i’ transforms an activation function. The activation unit’s value ‘P_i’ is computed as follows:

$${\mathrm{P}}_{\mathrm{i}}={\mathrm{f}(\mathrm{V}}_{\mathrm{i}})$$

(6)

4.2.2. ANN Performance

The performance of ANN is typically measured using metrics such as root mean square error (RMSE), absolute mean deviation (AMD) and the coefficient of determination (R²). These metrics are calculated as follows:

$${\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}^2=\left[{\displaystyle \frac{1}{\mathrm{n}}}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{M}\mathrm{i}-\mathrm{M}\mathrm{i}\mathrm{a})}^2\right]$$

(7)

$${\mathrm{R}}^2=1-{\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{M}\mathrm{i}-\mathrm{M}\mathrm{i}\mathrm{a})}^2}{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{(\mathrm{M}\mathrm{i}\mathrm{a}-\mathrm{M}\mathrm{m})}^2}}$$

(8)

$$\mathrm{A}\mathrm{M}\mathrm{D}=\left[{\displaystyle \frac{1}{\mathrm{n}}}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\displaystyle \frac{(\mathrm{M}\mathrm{i}-\mathrm{M}\mathrm{i}\mathrm{a})}{\mathrm{M}\mathrm{i}\mathrm{a}}}\right]\times 100$$

(9)

where ‘M_ia’ represents the actual value, ‘M_i’ denotes the predicted value, ‘M_m’ represents the mean of the actual values and ‘n’ indicates the number of data points.

4.3. Fuzzy ISM Approach

Initially introduced by Warfield (1974), ISM is used to analyse complex systems. It helps recognise fundamental interaction relationships among specific elements and construct a hierarchical structure model [35,38]. The ISM approach integrates several mathematical foundations, including digraph theory, expert elicitation methods, matrix operation theory and computer-aided calculations. In the traditional ISM approach, contextual relationships among system units are determined based on a binary relationship between each pair of units. This binary assumption overlooks the strength of these relationships and can influence expert judgment in identifying them. The fuzzy linguistic approach is integrated into traditional ISM to address these limitations. This enhancement improves rationality in determining contextual relationships among system units by expressing these relationships using fuzzy linguistic variables.

Triangular Fuzzy Linguistic Variables: This study introduces triangular fuzzy linguistic variables into ISM because they more effectively manage expert judgment information. The fuzzy ISM approach ensures a more nuanced and accurate representation of the relationships among system units, accommodating these relationships’ inherent uncertainties and strengths.

F-ISM Method Steps

Step 1: Pair-wise comparison of drivers. Triangular fuzzy numbers (TFNs) are employed in FISM. The linguistic variables used in FISM are detailed in

Table 3. Fuzzy scale for expert opinion [27].

Sr. No.	Linguistic Term	Corresponding TFN
1	No Influence (NI)	(0, 0.1, 0.3)
2	Low Influence (LI)	(0.1, 0.3, 0.5)
3	Average Influence (AI)	(0.3, 0.5, 0.7)
4	High Influence (HI)	(0.5, 0.7, 0.9)
5	Max Influence (MI)	(0.7, 0.9, 1)

. For the element (i and j) denoted by ‘HI’, the TFN is (0.5, 0.7 and 0.9). For the element (j and i), the TFN is (0, 0.1 and 0.3).

Table 4. Details of fuzzy hindrances between ith and jth.

Symbol	Element (ith to jth)	Element (jth to ith)	Symbol	Element (ith to jth)	Element (jth to ith)
V (MI)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	X (MI, HI)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)
V (HI)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	X (MI, AI)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)
V (AI)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	X (MI, LI)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)
V (LI)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	X (HI, MI)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)
A (MI)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	X (HI, AI)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)
A (HI)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	X (HI, LI)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)
A (AI)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	X (AI, MI)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)
A (LI)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	X (AI, HI)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)
X (MI)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	X (AI, LI)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)
X (HI)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	X (LI, MI)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)
X (AI)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	X (LI, HI)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)
X (LI)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	X (LI, AI)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)
O (NI)	(0, 0.1, 0.3)	(0, 0.1, 0.3)

lists the fuzzy influence between the ith and jth hindrances using fuzzy linguistic parameters. This section develops the structural self-interaction matrix (SSIM) by considering contextual relationships [35].

Step 2: Integration of expert opinions. Various strategies can be employed to integrate expert opinions, as advocated by [35]. The present study used the geometric mean approach to combine the experts’ opinions.

Step 3: Defuzzification. Many methods can be used for defuzzification, including the centre of gravity (COG), mean of maximum (MOM) and centre average methods. The COG method determines the value at the centre of the area under the curve, while the MOM approach identifies the point where balance is achieved. The COG method was employed in this research as per Equation (10). A TFN is defined as X = (f_i, g_i, h_i), where ‘f_i’ represents the smallest likely value, ‘g_i’ is the most probable value and ‘h_i’ is the highest possible value of any fuzzy event.

$${\mathsf{\lambda }}_{\mathrm{i}\mathrm{j}}={\displaystyle \frac{{\mathrm{f}}_{\mathrm{i}}+{\mathrm{g}}_{\mathrm{i}}+{\mathrm{h}}_{\mathrm{i}}}{3}}$$

(10)

Step 4: Partitioning the reachability matrix. A threshold (Ψ) based on the aggregate of expert opinions (from Step 3) is required. In this study, the average of all elements of the defuzzified matrix was calculated to determine the threshold. The reachability matrix was then constructed using Equations (11) and (12).

$${\mathrm{if}\mathsf{\lambda }}_{\mathrm{i}\mathrm{j}}\ge \Psi ⟶{\mathsf{\lambda }}_{\mathrm{i}\mathrm{j}}=1, {\mathsf{\lambda }}_{\mathrm{j}\mathrm{i}}=0$$

(11)

$${\mathrm{if}\mathsf{\lambda }}_{\mathrm{i}\mathrm{j}}\le \Psi ⟶{\mathsf{\lambda }}_{\mathrm{i}\mathrm{j}}=0, {\mathsf{\lambda }}_{\mathrm{j}\mathrm{i}}=1$$

(12)

Step 5: ‘Fuzzy Conical Matrix’. F-ISM is a more updated method than ISM. In ‘FISM’, the possibility of interaction is measured on a 0–1 scale, excluding 0 and 1. To achieve this, the reachability matrix must be raised to the power of ‘r+1’ to reach a steady state. This process is described in Equations (13) and (14).

$$\mathrm{M}=\mathrm{D}+\mathrm{I}$$

(13)

$$\mathrm{M}\mathrm{*}={\mathrm{M}}^{\mathrm{r}}={\mathrm{M}}^{\mathrm{r}+1}, \mathrm{r}>1$$

(14)

Here,

M: The reachability matrix represents the connections between elements in the system.

D: The direct relationship matrix shows direct connections between elements (initial direct relationship matrix D after defuzzification).

I: The identity matrix adds self-loops (each element is reachable from itself).

Step 6: Digraph development. An initial digraph that includes transitive links is constructed using the conical matrix. This initial digraph is represented by lines connecting edges (relationships) and nodes (elements). To achieve the final digraph, indirect links are removed. Positioning factors begin with placing the top-level factor at the highest position in the digraph. The second-level factor is positioned below it, continuing until the bottom-level factor is placed at the lowest in the digraph.

Step 7: The ‘MICMAC’ method creates a graph to classify factors based on their driving and dependence powers. This analysis is used to validate the interpretive structural model’s factors and derive results and conclusions. MICMAC analysis categorises the drivers into four groups: ‘linkage’, ‘autonomous’, ‘dependent’ and ‘independent’.

5. Results and Analysis

The results of our study are presented in this section. Here, we showcase the findings from EFA, ANN and fuzzy ISM.

5.1. Results Obtained through EFA

The EFA results reveal that the drivers of NZE in the MSC could be categorised into eight major groups, accounting for 67.725% of the total variance. The factor loadings for these drivers ranged from 0.603 to 0.906, exceeding the acceptable threshold. The final factor analysis results for net-zero adoption drivers are outlined in

Table 5. EFA results for net-zero adoption drivers.

Drivers (Predictor) Name	Sub-Drivers	Mean	SD	Item Loading	Eigenvalue	Cumulative %
Regulatory and Policy (RPP)	D1. Regulatory Compliance (RPP1)	3.360	1.380	0.725	3.741	14.383%
	D2. Government Incentives (RPP2)	3.227	1.342	0.906
	D3. Environmental Regulations (RPP3)	3.151	1.325	0.864
	D4. Industry Standards (RPP4)	3.093	1.411	0.682
Economic (EP)	D5. Cost Reduction (EP1)	3.233	1.407	0.815	3.175	26.600%
	D6. Financial Performance (EP2)	3.081	1.313	0.706
Technological (TP)	D7. Technological Advancements (TP1)	3.244	1.320	0.840	2.815	37.426%
	D8. Innovation Opportunities (TP2)	3.186	1.329	0.784
	D9. Eco-design (TP3)	3.279	1.206	0.765
Market and Consumer (MCP)	D10. Consumer Demand (MCP1)	3.378	1.276	0.868	2.087	45.454%
	D11. Brand Reputation (MCP2)	3.320	1.052	0.878
	D12. Competitive Advantage (MCP3)	3.512	1.046	0.794
	D13. Market Differentiation (MCP4)	3.157	1.249	0.780
Supply Chain and Procurement (SCPP)	D14. Supply Chain Pressure (SCPP1)	3.081	1.230	0.781	1.660	51.840%
	D15. Green Supply Chain Management (SCPP2)	3.169	0.962	0.871
	D16. Sustainable Procurement (SCPP3)	3.064	1.377	0.895
Corporate Governance and Ethics (CGEP)	D17. Corporate Social Responsibility (CGEP1)	3.169	1.087	0.829	1.424	57.317%
	D18. Long-term Viability (CGEP2)	3.477	1.152	0.807
	D19. Community Impact (CGEP3)	3.203	1.065	0.815
Environmental (EnP)	D20. Carbon Footprint Reduction (EnP1)	3.134	1.165	0.699	1.397	62.991%
	D21. Renewable Energy Adoption (EnP2)	3.128	1.158	0.804
	D22. Circular Economy Practices (EnP3)	3.017	1.221	0.833
	D23. Pollution Control (EnP4)	3.238	1.292	0.603
	D24. Climate Change Mitigation (EnP5)	3.238	1.132	0.649	1.309	67.757%
Operational (OP)	D25. Energy Efficiency (OP1)	3.355	1.319	0.726
		3.145	0.959	0.677
	D27. Product Lifecycle Management (OP3)	3.087	1.354	0.744

Kaiser–Meyer–Olkin measure of sampling (KMO): 0.841; Bartlett’s test of sphericity: Approx. Chi-square: 1760.33; df: 325; Sig.: 0.000. Rotation method: varimax with Kaiser normalisation.

.

5.2. Outcomes of ANN

In the second phase of the hybrid ANN-F-ISM MICMAC framework, we developed our ANN model using selected predictors. Fuzzy ISM effectively identifies interrelationships among factors impacting dependent variables, but traditional linear association methods, including ISM, fail to detect nonlinear relationships, leading to an overgeneralised view of complex human decision making [30,34]. We used ANN, a critical AI method, to address this and model the decision-making process. In this study, 70% of the data train the network model, while 30% test it, employing tenfold cross-validation. The ANN model’s input layer included eight variables: Regulatory and Policy, Economic, Technological, Market and Consumer, Supply Chain and Procurement, Corporate Governance and Ethics, Environmental and Operational. The dependent variable, net-zero adoption, is the output in the ANN model’s output layer. The evaluation results, indicated by RMSE, demonstrate the ANN model’s superior performance, as shown in

Table 6. Performance evaluation of the ANN model by RMSE.

NN	SS (Training)	SSE (Training)	SSE (Testing)	SS (Testing)	RMSE (Training)	RMSE (Testing)	TSS
ANN Config: 1	70	2.293	0.399	21	0.181	0.138	91
ANN Config: 2	66	2.113	0.453	25	0.179	0.135	91
ANN Config: 3	60	1.895	0.752	31	0.178	0.156	91
ANN Config: 4	56	1.732	0.212	35	0.176	0.078	91
ANN Config: 5	70	3.058	0.121	21	0.209	0.076	91
ANN Config: 6	65	1.226	0.164	26	0.137	0.079	91
ANN Config: 7	67	1.391	0.171	24	0.144	0.084	91
ANN Config: 8	58	1.379	0.012	33	0.154	0.019	91
ANN Config: 9	53	1.091	0.056	38	0.143	0.038	91
ANN Config: 10	61	1.19	0.014	30	0.140	0.022	91
Mean		1.74	0.24		0.16	0.08
SD		0.59	0.22		0.02	0.05

Config = Configuration, SSE = Sum of Square Error, NN = Neural Network, SS = Sample Size and TSS = Total Sample Size.

.

The mean RMSE value of the testing model stands at 0.08, while that of the training model is 0.16, indicating the superior performance of the NN model in capturing nonlinear relationships among independent variables influencing net-zero adoption in the MSC. The proposed framework of ANN is illustrated in Figure 8. A sensitivity analysis of the NN model reveals the relative importance of independent variables in predicting net-zero adoption. The normalised importance quantifies how changes in independent variables affect the model-predicted outcomes.

According to the ANN findings, Environmental drivers emerge as the most influential independent variable, attributing 100% importance to the prediction of net-zero adoption. Following closely are Economic drivers at 60.38%, Technological drivers at 59.05%, Market and Consumer at 47.02%, Corporate Governance and Ethics at 46.22%, Regulatory and Policy at 43.94%, Operational drivers at 33.75% and Supply Chain and Procurement at 24.85%. A comprehensive overview of RMSE is represented in Figure 9. This paragraph maintains the structure and content of the original while adapting it to fit the context of drivers influencing net-zero adoption in the MSC.

To measure the predictive power of each input neuron, a sensitivity analysis (

Table 7. Summary of normalised importance (Sensitivity analysis).

				Independent Variable Importance
Predictor Name	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	NI (in %)	Avg. Score	NS (%)	Rank
RPP	21.79	9.00	67.11	59.54	46.71	63.18	24.80	68.50	33.80	30.23	0.42	43.94	6
EP	37.86	76.64	46.84	53.62	73.46	13.09	46.35	88.01	47.72	100.00	0.58	60.38	2
TP	33.54	97.40	55.65	68.04	73.67	39.31	49.25	59.72	63.40	30.72	0.57	59.05	3
MCP	20.27	24.05	27.80	73.95	40.29	30.71	54.23	97.83	23.83	61.51	0.45	47.02	4
SCPP	2.81	23.51	61.67	17.36	38.01	16.60	11.07	34.62	21.82	12.72	0.24	24.85	8
EnP	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	66.54	0.97	100.00	1
CGEP	39.13	43.36	64.76	46.55	27.10	33.27	60.19	59.56	41.37	31.42	0.45	46.22	5
OP	29.56	39.55	62.51	32.30	26.55	41.14	20.61	30.88	14.94	28.19	0.33	33.75	7

NI = Normalised Importance, NS = Normalise Score and Avg. = Average.

) was conducted to obtain the normalised importance of these neurons. This was performed by dividing their relative importance by the maximum significance and presenting it as a percentage [32,36]. The results show that Environmental drivers are the most critical predictor with a score of 100%, followed by EP, which has a normalised importance of 60.38%, and TP, which has a score of 59.05%.

5.3. Outcomes through F-ISM MICMAC

In this section, we present the outcomes related to RQ3. Initially, we identified 30 potential drivers related to net-zero adoption, as represented in Figure 4. For further analysis using fuzzy ISM techniques, we only considered 27 drivers that were distinctly agreed upon by most experts. For the F-ISM method, we selected a panel of six experts, ensuring a balance between academia and industry to minimise individual bias. To establish pair-wise associations among the 27 drivers, the experts were asked to provide their opinions on the relationships across the rows and columns using four options: ‘achieved by’, ‘leads to’, ‘bidirectional’ or ‘no relation’.

Formulation of SSIM

A pair-wise association among the finalised 27 drivers is developed in the second step. An SSIM is then established based on the initial associations among the drivers across the rows and columns of

Table A1. Structural self-interaction matrix (SSIM).

Drivers

D27

D26

D25

D24

D23

D22

D21

D20

D19

D18

D17

D16

D15

D14

D13

D12

D11

D10

D9

D8

D7

D6

D5

D4

D3

D2

D1

D1

V (AI)

V (LI)

X (HI, AI)

A (AI)

X (AI)

V (AI)

A (MI)

V (AI)

X (MI)

X (LI, HI)

X (MI)

V (MI)

X (MI)

V (MI)

A (MI)

V (AI)

X (MI)

X (LI, HI)

V (AI)

X (HI, AI)

V (AI)

X (MI)

A (MI)

X (LI, AI)

X (AI, MI)

X (HI, AI)

1

D2

X (AI)

V (LI)

X (AI)

X (LI)

X (AI, MI)

X (MI)

X (LI, HI)

X (HI)

A (MI)

X (HI)

A (MI)

X (MI, LI)

X (LI, HI)

O (NI)

X (AI, MI)

X (LI, HI)

A (MI)

A (MI)

X (HI)

A (MI)

X (LI, AI)

V (MI)

X (HI)

A (MI)

X (HI)

1

D3

O (NI)

A (AI)

A (LI)

X (HI)

V (LI)

X (MI, LI)

A (MI)

V (HI)

X (MI, LI)

X (HI, MI)

V (HI)

O (NI)

A (MI)

X (LI, HI)

X (HI)

X (HI, MI)

X (MI, LI)

X (LI, AI)

X (HI)

X (LI, AI)

O (NI)

X (HI, AI)

V (MI)

X (MI, LI)

1

D4

A (MI)

V (AI)

X (HI)

X (MI, LI)

X (HI, MI)

O (NI)

X (HI, LI)

X (MI, HI)

V (MI)

X (LI, HI)

V (MI)

X (LI, HI)

V (MI)

X (AI, MI)

X (MI)

A (MI)

V (MI)

V (MI)

X (HI, AI)

X (MI)

X (AI, MI)

X (LI, HI)

X (LI, HI)

1

D5

X (MI)

A (LI)

X (HI, LI)

X (MI)

V (MI)

X (HI, MI)

X (MI, LI)

A (MI)

X (MI)

X (HI, LI)

A (MI)

X (MI, LI)

X (HI, LI)

A (MI)

V (AI)

V (MI)

X (MI)

X (LI, HI)

X (HI)

X (LI, HI)

X (MI, LI)

A (MI)

1

D6

V (MI)

A (MI)

X (MI, LI)

X (HI, LI)

X (MI, HI)

A (MI)

V (MI)

V (AI)

X (MI, LI)

X (HI, AI)

X (HI)

X (HI, MI)

X (HI, AI)

X (LI, AI)

X (LI)

X (LI, AI)

X (HI, LI)

X (MI, HI)

X (LI, HI)

X (AI, MI)

X (AI, MI)

1

D7

X (HI)

X (MI, HI)

V (HI)

X (HI, MI)

V (MI)

X (HI, LI)

A (MI)

X (LI)

X (LI, AI)

X (MI)

X (LI)

V (MI)

X (HI, LI)

X (MI)

X (MI, HI)

X (HI, AI)

V (MI)

X (HI, LI)

A (MI)

A (MI)

1

D8

A (MI)

A (MI)

V (AI)

V (MI)

A (MI)

O (NI)

X (HI, AI)

X (HI, MI)

O (NI)

A (MI)

X (LI, HI)

X (LI, AI)

X (AI, MI)

X (HI, LI)

X (LI)

X (HI)

X (LI, HI)

X (HI)

X (HI)

1

D9

X (MI, HI)

V (MI)

V (MI)

X (MI, LI)

X (MI)

X (MI)

X (LI)

X (HI, LI)

X (LI, HI)

X (HI, LI)

X (MI, HI)

X (HI, LI)

X (HI, LI)

X (MI)

X (MI, LI)

X (LI)

X (MI, HI)

X (AI, MI)

1

D10

X (MI)

A (MI)

X (MI, LI)

X (HI, LI)

X (MI, HI)

X (LI, HI)

V (MI)

X (MI, HI)

X (HI, LI)

V (AI)

V (AI)

A (LI)

X (AI, MI)

X (MI, HI)

X (HI)

V (AI)

V (MI)

1

D11

V (MI)

X (HI, MI)

A (MI)

X (LI)

A (LI)

V (AI)

V (AI)

V (AI)

X (MI, LI)

V (AI)

X (HI)

X (HI)

X (MI)

V (AI)

V (HI)

V (MI)

1

D12

V (HI)

V (HI)

A (LI)

X (HI, MI)

X (HI)

A (MI)

V (HI)

X (HI)

X (MI, HI)

A (LI)

V (HI)

X (LI, HI)

V (HI)

A (MI)

V (MI)

1

D13

V (AI)

X (LI)

V (AI)

V (AI)

O (NI)

X (LI)

X (LI, AI)

V (MI)

V (HI)

X (MI)

X (LI, HI)

O (NI)

O (NI)

V (MI)

1

D14

A (MI)

V (MI)

X (MI, LI)

X (MI)

V (AI)

X (MI, HI)

O (NI)

X (LI, HI)

V (AI)

A (MI)

O (NI)

X (LI, HI)

X (MI)

1

D15

X (MI)

X (AI, MI)

X (LI, MI)

X (MI, HI)

X (MI)

O (NI)

X (LI, MI)

X (MI, HI)

X (LI)

X (LI, MI)

O (NI)

X (MI)

1

D16

V (MI)

V (AI)

X (HI)

A (MI)

X (HI, LI)

X (LI, MI)

X (LI)

O (NI)

X (HI, AI)

O (NI)

X (MI)

1

D17

A (MI)

V (AI)

O (NI)

V (MI)

O (NI)

V (AI)

X (LI, HI)

V (AI)

X (LI, MI)

X (MI)

1

D18

V (HI)

X (HI, MI)

X (MI)

X (LI)

X (MI)

O (NI)

O (NI)

X (MI, HI)

X (LI, MI)

1

D19

A (LI)

X (MI)

X (MI, HI)

V (AI)

X (MI, HI)

O (NI)

X (MI)

X (LI, MI)

1

D20

X (LI)

A (MI)

O (NI)

X (MI)

O (NI)

X (MI)

X (MI, HI)

1

D21

O (NI)

X (LI, MI)

X (MI, LI)

O (NI)

V (AI)

X (MI)

1

D22

A (MI)

X (LI)

V (MI)

X (MI)

O (NI)

1

D23

A (MI)

X (MI, HI)

X (MI)

X (LI, MI)

1

D24

X (MI)

X (LI, MI)

V (MI)

1

D25

V (AI)

X (MI)

1

D26

A (MI)

1

D27

1

(refer to Appendix A), with ‘i’ representing the row number and ‘j’ representing the column number. The symbols O, X, V and A represent the correlation paths among drivers. Specifically:

• O indicates no relation between the drivers.
• X signifies a bi-directional association between the drivers.
• V means driver ‘i’ helps to achieve driver ‘j’.
• A implies driver ‘j’ helps to achieve driver ‘i’.

An SSIM matrix is constructed based on the interrelationships among the drivers and is reviewed in collaboration with experts in the rubber and tire manufacturing industries.

Fuzzy calculations were employed to use the aggregated SSIM and construct a reachability matrix. As shown in

Table A2. Fuzzy reachability matrix.

Drivers	D27	D26	D25	D24	D23	D22	D21	D20	D19	D18	D17	D16	D15	D14	D13	D12	D11	D10	D9	D8	D7	D6	D5	D4	D3	D2	D1
D1	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	1, 1, 1
D2	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	1, 1, 1	(0.3, 0.5, 0.7)
D3	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	1, 1, 1	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)
D4	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	1, 1, 1	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)
D5	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	1, 1, 1	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)
D6	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	1, 1, 1	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.7, 0.9, 1)
D7	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	1, 1, 1	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)
D8	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	1, 1, 1	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)
D9	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	1, 1, 1	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)
D10	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)
D11	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	1, 1, 1	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)
D12	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	1, 1, 1	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)
D13	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	1, 1, 1	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)
D14	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	1, 1, 1	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)
D15	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)
D16	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)
D17	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)
D18	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	1, 1, 1	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)
D19	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	1, 1, 1	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)
D20	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)
D21	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	1, 1, 1	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)
D22	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	1, 1, 1	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)
D23	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	1, 1, 1	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)
D24	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	1, 1, 1	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)
D25	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	1, 1, 1	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.3, 0.5, 0.7)	(0.3, 0.5, 0.7)
D26	(0, 0.1, 0.3)	1, 1, 1	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.1, 0.3, 0.5)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)	(0, 0.1, 0.3)
D27	1, 1, 1	(0.7, 0.9, 1)	(0.3, 0.5, 0.7)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.1, 0.3, 0.5)	(0.1, 0.3, 0.5)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0.7, 0.9, 1)	(0.5, 0.7, 0.9)	(0, 0.1, 0.3)	(0.7, 0.9, 1)	(0.7, 0.9, 1)	(0, 0.1, 0.3)	(0.3, 0.5, 0.7)	(0, 0.1, 0.3)

(refer to Appendix A), the fuzzy reachability matrix is derived. Subsequently, the FISM hierarchical diagram, based on the final reachability matrix (FRM) (see

Table 8. Final reachability matrix (FRM).

Drivers	D1	D2	D3	D4	D5	D6	D7	D8	D9	D10	D11	D12	D13	D14	D15	D16	D17	D18	D19	D20	D21	D22	D23	D24	D25	D26	D27	Driving Power
D1	1	1	0	0	0	1	0	1	0	0	1	0	0	1	1	1	1	0	1	0	0	0	0	0	1	0	0	11
D2	0	1	1	0	1	1	0	0	1	0	0	0	1	0	1	1	0	0	0	0	0	1	0	0	0	0	0	9
D3	1	1	1	1	1	1	1	0	0	0	1	1	1	0	1	0	1	0	1	1	0	0	0	1	0	0	0	15
D4	1	1	0	1	0	1	0	1	1	1	1	0	1	0	1	0	1	0	1	1	1	1	0	1	1	0	0	17
D5	1	1	0	1	1	0	1	0	1	0	1	1	0	0	1	1	1	0	0	0	1	1	0	1	0	0	1	15
D6	0	1	0	0	1	1	0	0	0	1	1	0	1	0	1	1	1	0	1	0	0	0	1	0	1	1	1	14
D7	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	0	1	1	1	1	1	1	26
D8	0	1	1	1	1	1	1	1	1	1	0	1	1	1	0	0	1	0	0	0	1	1	1	1	0	1	1	19
D9	1	0	0	0	0	1	1	0	1	0	1	0	1	1	1	1	1	1	0	1	0	1	1	1	1	1	1	18
D10	1	1	1	0	1	0	1	0	0	1	1	0	0	1	0	0	0	0	1	0	0	0	1	0	0	0	0	10
D11	1	1	0	1	0	0	0	1	0	0	1	1	1	0	1	0	0	0	1	0	0	0	0	0	0	0	0	9
D12	1	1	0	0	1	0	1	0	0	0	0	1	1	0	1	0	1	1	1	1	1	1	1	0	1	1	1	17
D13	1	1	1	1	0	0	1	0	0	1	0	0	1	1	0	1	1	1	0	1	0	1	1	1	1	1	1	18
D14	0	1	1	0	1	1	1	0	0	0	1	1	0	1	1	1	0	0	0	0	1	0	0	1	1	1	0	14
D15	0	1	1	0	0	0	1	0	0	1	0	0	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	19
D16	0	1	1	1	1	0	0	0	0	1	0	1	1	1	0	1	1	0	1	0	1	1	0	1	1	1	1	17
D17	0	1	0	0	1	0	1	0	0	1	0	0	1	1	1	0	1	1	0	0	0	0	0	1	0	0	0	10
D18	1	0	0	1	1	0	0	1	0	1	1	1	0	1	1	1	0	1	0	1	0	0	1	0	1	1	1	16
D19	0	1	0	0	0	0	1	1	1	0	0	1	1	0	1	0	0	1	1	1	1	0	1	0	1	0	1	14
D20	1	1	0	0	1	1	1	0	0	0	1	0	0	1	0	1	0	0	1	1	0	1	0	0	0	0	0	11
D21	1	1	1	0	0	0	1	1	1	0	0	0	1	0	1	0	0	1	0	0	1	1	0	0	1	0	0	12
D22	1	1	0	1	1	0	0	1	0	1	1	1	1	1	0	1	1	1	0	0	0	1	0	1	0	0	0	15
D23	1	1	1	1	0	0	0	0	0	1	0	1	0	0	0	0	0	0	0	0	1	0	1	0	1	1	1	11
D24	1	1	1	0	1	1	1	1	0	1	0	1	0	1	0	1	0	1	1	1	1	1	1	1	1	1	1	21
D25	0	1	1	0	1	0	1	0	0	0	1	0	1	0	1	0	1	0	0	1	0	1	1	0	1	1	0	13
D26	1	1	0	0	1	1	1	0	1	1	1	1	0	0	0	0	0	0	0	0	0	1	1	0	0	1	0	12
D27	1	1	1	1	1	0	1	0	0	0	0	0	0	1	1	0	1	0	1	1	1	1	1	1	1	1	1	18
Dependence	18	25	14	12	18	12	18	10	9	14	15	14	17	15	18	14	16	11	14	13	12	17	15	14	17	15	14

) and level partitioning of net-zero adoption drivers (see

Table 9. Level partitioning of net-zero adoption drivers.

Drivers	Reachability Set	Antecedent Set	Intersection Set	Level
D1	(2, 15, 18)	(2, 15, 18, 22, 26)	(2, 15, 18)	IV
D2	(1, 12, 14)	(1, 12, 19, 27)	(1, 12)	VII
D3	(1, 2, 4, 5, 12, 17, 23, 24)	(1, 4, 5, 12, 17, 23, 24, 27)	(1, 4, 5, 12, 17, 23, 24)	I
D4	(4, 6, 7, 11, 19, 26)	(4, 6, 7, 8, 11, 18, 19, 21, 26, 27)	(4, 6, 11, 19, 26)	III
D5	(1, 11, 12, 15, 16, 23, 27)	(2, 4, 11, 12, 13, 14, 15, 17, 23, 24, 25)	(1, 11, 12, 15, 23)	VIII
D6	(1, 2, 4, 7, 9, 10, 16, 20, 22, 23)	(1, 2, 4, 7, 8, 9, 11, 14, 15, 19, 20, 23, 24)	(1, 2, 4, 7, 9, 20, 23)	I
D7	(1, 3, 5, 6, 12, 15, 17, 18, 20, 21, 26, 27)	(1, 2, 3, 15, 17, 18, 20, 24, 25, 26, 27)	(1, 3, 15, 17, 18, 20, 26, 27)	V
D8	(1, 11, 12, 15, 17, 18, 20)	(1, 2, 3, 7, 8, 9, 11, 14, 15, 19, 20, 23, 24)	(1, 11, 15, 20)	VII
D9	(1, 5, 6, 12, 15, 17, 18, 23)	(3, 4, 8, 12, 14, 15, 17, 20, 23, 24, 25, 26)	(12, 15, 17, 23)	I
D10	(2, 3, 4, 5, 6, 8, 15, 17, 18, 20, 21)	(1, 2, 3, 7, 8, 9, 11, 14, 19, 21, 24)	(2, 3, 8, 21)	IV
D11	(2, 3, 5, 7, 12, 15, 17, 18, 20)	(1, 2, 3, 7, 8, 9, 11, 14, 15, 19, 20, 23, 24)	(2, 3, 7, 15, 20)	VII
D12	(2, 5, 18, 19)	(2, 3, 7, 8, 9, 11, 18, 19, 20, 23, 24)	(2, 5, 18, 19)	III
D13	(6, 18, 21, 23, 24, 26)	(8, 9, 11, 14, 15, 19, 20, 23, 24, 26)	(6, 23, 24, 26)	V
D14	(1, 10, 17, 18, 24)	(1, 2, 10, 15, 19, 20, 23, 24)	(1, 10, 24)	II
D15	(4, 5, 6, 12, 15, 17, 18, 25, 27)	(2, 3, 7, 8, 9, 12, 14, 25, 23, 27)	(12, 25, 27)	VI
D16	(8, 9, 18, 21)	(1, 2, 3, 7, 8, 9, 11, 14, 15, 18, 23, 24, 25)	(8, 9, 18)	VIII
D17	(1, 2, 5, 6, 12, 16, 19, 20)	(1, 2, 3, 5, 6, 9, 11, 14, 16, 19, 20, 23, 24)	(1, 2, 5, 6, 16, 19, 20)	I
D18	(1, 3, 5, 6, 12, 15, 17, 18, 20)	(2, 3, 7, 8, 9, 11, 12, 15, 18, 20, 23)	(3, 12, 15, 18, 20)	VI
D19	(2, 3, 5, 6, 12, 15, 17, 18, 20)	(2, 3, 7, 8, 9, 11, 14, 15, 19, 20, 23, 24)	(2, 15, 20)	II
D20	(1, 2, 3, 5, 15, 17, 18, 21, 22)	(1, 7, 8, 9, 11, 14, 15, 17, 19, 20, 12, 23, 24)	(1, 15, 17)	V
D21	(4, 6, 12, 17, 18, 20)	(1, 2, 4, 6, 8, 9, 11, 14, 15, 19, 20, 23, 24)	(4, 6, 20)	VII
D22	(1, 5, 6, 14, 18, 20, 26)	(1, 2, 8, 9, 11, 14, 15, 20, 23, 24, 26)	(1, 14, 20, 26)	VI
D23	(1, 6, 12, 15, 17, 18)	(1, 8, 11, 14, 15, 18, 20, 23)	(1, 15, 18)	III
D24	(3, 5, 6, 12, 15, 17, 18, 21, 26)	(1, 2, 3, 7, 8, 9, 11, 12, 15, 19, 20, 23, 24, 26)	(3, 12, 15, 26)	IV
D25	(1, 3, 5, 6, 12, 22, 24)	(1, 2, 3, 7, 8, 9, 11, 14, 15, 19, 20, 22, 27)	(1, 3, 22, 27)	VI
D26	(18, 21, 23)	(18, 20)	(18)	IX
D27	(2, 9, 13, 16, 20, 27)	(1, 2, 3, 7, 8, 9, 13, 14, 15, 19, 20, 22, 27)	(2, 9, 13, 16, 20, 27)	II

), is presented. Figure 10 illustrates the FISM hierarchy. As indicated in the F-MICMAC results (Figure 11), the driving power and dependent quantities from the FRM were used to categorise the drivers into four groups. MICMAC analysis primarily aims to estimate drivers’ drive and dependence powers. The drivers are classified into four groups:

Autonomous drivers: These drivers have weak drive and dependence and are placed in the first cluster. They need to be connected to the system, with few but solid links.

Dependent drivers: The second cluster includes drivers with strong dependence power but weak drive power.

Linkage drivers: This cluster consists of drivers with solid dependence and strong drive power. They are unstable, and any action on them will impact others and have a negative effect on themselves.

Independent drivers: The fourth cluster contains drivers with solid drive power but weak dependence power. A driver with extreme drive power called the ‘key driver’, falls into the independent or linkage hindrance category.

6. Findings and Discussion

Following the objective of this study outlined in the introduction, an SLR was conducted and identified 30 potential drivers of net-zero adoption in the MSC. We conducted a comprehensive questionnaire survey and selected a panel of seven experts based on their knowledge and experience. This panel identified 27 out of 30 drivers as distinct and crucial. Valid responses from 173 manufacturing sectors were analysed using EFA. Subsequently, a specific manufacturing industry was selected for a case study to test the proposed framework. An integrated decision support system, incorporating ANN and fuzzy ISM MICMAC, was utilised to prioritise, analyse the influence, identify the interaction relationship among the drivers, construct a hierarchical structure among these identified drivers and validate the framework. These 27 distinct drivers, agreed upon by most experts, were subjected to further analysis using the ANN-fuzzy ISM MICMAC approach.

The EFA approach was utilised to categorise the identified drivers [41], while NN was employed to identify ‘non-linear relationships’ and prioritise. The F-ISM approach was then applied to establish the interrelationships among the drivers supported by ANN and to determine the critical drivers based on their driving power.

The ANN results indicate that the most critical driver (predictor) groups are Environmental drivers, which emerge as the most influential independent variable, attributing 100% importance to the prediction of net-zero adoption. Following closely are Economic drivers at 60.38%, Technological drivers at 59.05%, Market and Consumer at 47.02%, Corporate Governance and Ethics at 46.22%, Regulatory and Policy at 43.94%, Operational drivers at 33.75% and Supply Chain and Procurement at 24.85%. A study emphasised the importance of environmental drivers in adopting the net-zero concept in the MSC [27]. The ‘RMSE’ value of the testing model stands at 0.08, while for the training model, it is 0.16, indicating the superior performance of the ANN model in capturing nonlinear relationships among independent variables influencing net-zero adoption in the MSC. A sensitivity analysis is conducted to identify the relative importance of the drivers of net-zero adoption.

To establish a pair-wise comparison between the 27 drivers, experts were asked to evaluate the relationships between rows and columns using four options: ‘achieved by’, ‘leads to’, ‘bidirectional’ and ‘no relation’. The resulting F-ISM-based structural model, represented in Figure 10, arranges these drivers into nine distinct levels within the hierarchy. Figure 10 represents the F-ISM model of net-zero adoption drivers to MSC, organised into nine hierarchical layers. The lowest levels are essential, impacting higher-level drivers in the hierarchy. The most important drivers, found in levels VII, VIII and IX, include five key drivers: D4, D6, D8, D9 and D18. Policymakers and decision makers should prioritise addressing these drivers to ensure long-term success in the industry. The moderately important drivers, spanning levels IV, V and VI, consist of 13. Those above and below them influence these drivers in the hierarchy. The least essential drivers are found in levels I, II and III; nine are in this group. These drivers are less substantial but still relevant to the overall structure.

The F-MICMAC analysis is classified into four groups: autonomous group drivers, dependent group drivers, linkage group drivers and driving group drivers.

Autonomous group drivers: The autonomous group drivers are marked by a lack of ‘influential power’ and ‘dependency’, requiring minimal attention by ‘decision-makers’. Only one driver, D21, belongs to the autonomous group of drivers.

Dependent group drivers: This group comprises drivers with low influential power but high dependency. Eight drivers fall into this group: D1, D2, D10, D11, D17, D20, D23 and D26.

Linkage group drivers: Drivers in this group have high driving power and dependency. Out of the 27 drivers, 13 belong to this group, as represented in Figure 11.

Driving group drivers: This group comprises the most crucial drivers with high influence and low dependence power. Five drivers belong to this group: D4, D6, D8, D9 and D18.

The outcomes of this study align with past research on MSC to achieve sustainability, such as [42], which identified technological innovation and regulatory frameworks as key drivers for sustainability. Our hybrid approach extends these insights by quantifying the influence of these drivers and elucidating their interrelationships.

Past studies have also emphasised the importance of circular procurement practices and economic viability in decarbonisation efforts [5]. Our results support these findings and highlight how these factors operate within the Indian manufacturing sector. The use of ANN and the fuzzy ISM approach in our study provides robust validation, underscoring their strategic significance in planning decarbonisation initiatives.

This discussion integrates past study findings, emphasising the implications of our current research. It underscores the necessity for continuous innovation and the adoption of advanced technologies like IoT and blockchain to enhance supply chain transparency and efficiency. These insights validate our results and highlight the practical applications for industry practitioners and policymakers aiming to achieve NZE in the manufacturing sector.

7. Conclusions, Contributions and Recommendations for Future Research

This study proposes conducting net-zero adoption in the MSC. Only a few studies have examined drivers affecting net-zero adoption but have yet to explore their interrelationships. An integrated two-stage ANN-FISM approach was developed to address this gap and analyse drivers influencing net-zero adoption for the MSC. We identified 30 potential drivers through an SLR, and a panel of seven experts narrowed this to 27 distinct drivers for further analysis. These drivers were further distributed into eight main groups (predictors). The ANN analysis quantified the importance of these drivers, with Environmental drivers being the most influential at 100%, followed by Economic drivers at 60.38% and Technological drivers at 59.05%.

The F-ISM MICMAC analysis categorized the 27 drivers into four quadrants: 1 autonomous, 8 dependent, 13 linkage and 5 driving-group drivers, highlighting the complex interdependencies that shape net-zero adoption in MSC.

The hybrid approach, combining ANN for predicting key drivers and fuzzy ISM for uncovering their relationships, provides a comprehensive framework for understanding net-zero adoption. The findings offer valuable insights for the Indian manufacturing industries, guiding the development of effective business policies and strategies for enhancing performance through net-zero practices. The model was validated by industry experts, with all hypotheses confirmed, reinforcing the robustness of the proposed methodology.

7.1. Theoretical Contributions

Previous reports have highlighted the growing interest in NZE and its effect on environmental sustainability. Still, more case studies, reviews and reports are needed. In developing countries such as India, net zero is a new concept that requires more in-depth analysis and vision. Through SLR, we identified 27 potential drivers related to net-zero adoption that affect the MSC. To validate these drivers, 173 responses were collected and analysed using EFA. We conducted a case study in a specific manufacturing industry. The ANN technique is applied to weigh the drivers’ nonlinear effects. The results reveal that Environmental drivers, which emerge as the most influential independent variable, attribute 100% importance to the prediction of net-zero adoption. Following closely are Economic drivers at 60.38% and Technological drivers at 59.05%, which are particularly important for the MSC. The Indian manufacturing sector must focus on R&D practices and establish dedicated teams for net-zero initiatives. Collaboration with research and academic institutions will allow manufacturing industries to share their views and experiences on achieving net zero. Embracing these drivers will help the MSC reach the net-zero goal. Investing in green initiatives and renewable infrastructure will help the manufacturing sector achieve its net-zero aim.

7.2. Managerial Implications

The present study provides valuable insights into the drivers influencing net-zero adoption and their effects on business performance, which can benefit managers in various sectors, including service, manufacturing and process industries. Effective strategies can be formulated for adopting the net-zero concept in their industries and evaluating the time required for its efficient implementation. Since ‘Environmental drivers’ were identified as the most crucial drivers in adopting net zero in the MSC, organisations must incorporate trust-building measures into their business strategies. Regular assessments should ensure trust in adopting new technology among managers. Feedback forms, surveys and interviews can help gauge managers’ trust. Maintaining Environmental drivers is crucial for successfully adopting net zero in the MSC. Therefore, organisations need to ensure that Environmental drivers are continuously upheld. This necessitates periodic revisions of their strategies to embrace new technologies like net zero effectively.

7.3. Limitations

This study is limited to the Indian manufacturing industry, which may restrict the applicability of findings to other regions. Future studies should include data from diverse countries to enhance global relevance and provide cross-cultural insights into achieving NZE in MSC. Secondly, the data collection method, which relies heavily on expert opinions, may introduce bias. To mitigate this, future studies should incorporate a broader range of decision-making approaches, including surveys and interviews with a diverse sample of stakeholders. Lastly, while the ANN-F-ISM approach employed in this study is robust, additional validation tools such as SEM could further corroborate the findings. Employing different analytical techniques in future research would provide a more comprehensive understanding of the drivers influencing net-zero adoption in the MSC. These limitations highlight the need for further studies to enhance the robustness and generalisability of the findings across various emerging economies and industry contexts.