Development of a framework for sustainability assessment of the machining process through machining parameter optimisation technique

ABSTRACT

This study introduces a comprehensive framework that integrates Multi-objective optimisation with Multiple-Criteria Decision Making (MCDM) techniques to enhance decision-making in sustainable machining processes. Through a systematic literature review, the framework is developed, incorporating sustainability indicators, experimental validation, and multi-objective optimisation algorithms to identify the optimal machining conditions. To address the challenge of subjective judgements and the multiple solutions generated by multi-objective optimisation, MCDM techniques are employed to select the best solution. By considering economic viability, environmental impact reduction, and social commitment, the proposed framework overcomes the limitations of traditional optimisation approaches. It provides an objective and holistic approach to decision-making, promoting sustainability in machining processes. The framework’s effectiveness and applicability are demonstrated through a case study, validating its potential for practical implementation and informed decision-making

Graphical abstract

KEYWORDS:

• Machine parameter optimisation
• multi-objective optimisation
• sustainability assessment
• multi-criteria decision-making
• validation
• sustainable manufacturing

1. Introduction

The manufacturing industry plays a significant role in environmental degradation (Rajemi, Mativenga, and Aramcharoen 2010), with machines such as milling, lathe, and EDM being major energy consumers, accounting for up to 50% of total energy consumption (Hu et al. 2019). To address this issue, improving the sustainability of manufacturing processes has become crucial. Machine parameter optimisation has emerged as a promising approach in recent years, aiming to optimise machining parameters to achieve desired performance while minimising environmental impact (Cai et al. 2022). This optimisation technique has shown considerable improvements in machining performance, including reduced tool wear and increased material removal rates, while also reducing energy consumption and carbon emissions (Pawanr, Kant Garg, and Routroy 2019).

Machine parameter optimisation is a multi-objective problem, where multiple objectives may conflict with each other, creating trade-offs that need to be considered for determining the best solution. It is important to take into account these trade-offs and consider the optimisation outcomes in a multi-objective context. By employing multi-objective optimisation techniques, a set of optimal solutions can be obtained, each representing different trade-offs between the objectives. The selection of the best solution from this set requires a comprehensive evaluation that considers the inherent trade-offs and takes into account the specific requirements and priorities of the problem at hand. Some of the examples of these tradeoffs are given below:

• Surface Finish vs. Material Removal Rate: Achieving a smooth and high-quality surface finish often requires using lower cutting speeds and feeds, which may result in a slower material removal rate. Conversely, maximising material removal rate may lead to rougher surface finishes. Balancing these objectives is crucial to achieve the desired level of surface finish while maintaining productivity.

• Tool Life vs. Productivity: Increasing cutting speeds and feeds can enhance productivity by reducing machining time. However, this may also result in increased tool wear and shorter tool life. Optimising machine parameters to extend tool life while maintaining acceptable productivity levels requires finding an optimal balance between these conflicting objectives.

• Energy Consumption vs. Machining Time: Minimising energy consumption is an essential aspect of sustainability. However, reducing cutting speeds and feeds to lower energy consumption may lead to longer machining times, which can be detrimental to overall productivity. Finding the optimal parameters that strike a balance between energy efficiency and machining time is crucial for sustainable machining operations.

• Dimensional Accuracy vs. Tool Wear: Achieving high dimensional accuracy requires precise control over machining parameters. However, aggressive cutting conditions that ensure dimensional accuracy can accelerate tool wear, leading to increased costs associated with tool replacement and maintenance. Striking a balance between dimensional accuracy and tool wear is necessary to optimise both quality and cost-effectiveness.

Multi-objective optimisation techniques offer several benefits. They facilitate improved decision-making by providing a comprehensive understanding of the trade-offs between different objectives. The outcome of multi-objective optimisation is a set of solutions known as the Pareto front, representing the best trade-offs between conflicting objectives. This provides decision-makers with a range of options to choose from, enhancing flexibility and adaptability. Moreover, by incorporating multiple sustainability objectives, such as energy consumption and tool wear, multi-objective optimisation enables the development of more sustainable machining processes.

However, one limitation of multi-objective optimisation is that it yields multiple solutions, and the selection of the final solution depends on user preferences and experience. This process often involves subjective judgements and trade-offs based on individual expertise and domain knowledge. To overcome this limitation, this study proposes a framework that combines multi-objective optimisation with Multiple-Criteria Decision Making (MCDM) techniques. MCDM techniques, such as Grey Relational Analysis (GRA), Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), and Analytic Hierarchy Process (AHP), can be employed to select the best solution from the set of optimal solutions provided by multi-objective optimisation. These techniques consider different criteria and assign weights to each criterion, helping to quantify preferences. Therefore, the aim of this study is to propose a framework to improve decision-making in the machining process, the objectives of the framework are as follows:

• Design and conduct experiments to collect and calculate sustainability indicators machining process under different machining conditions.

• Develop a mathematical model that captures the relationships between the machining conditions and the response variables.

• Apply a multi-objective optimisation algorithm to the mathematical model to find the optimal machining conditions that minimise or maximise respective sustainability indicators, thereby improving economic viability and environmental impact.

• Evaluate the sustainability of the machining process by analysing the results of the multi-objective optimisation through MCDM.

By addressing these objectives, this research aims to enhance the sustainability of magnesium alloy machining, improving economic feasibility and reducing environmental impact.

2. Literature review

2.1. Sustainability in manufacturing processes

The existing literature on sustainability assessment in manufacturing processes has predominantly focused on product-level sustainability rather than process-level evaluation (Abubakr et al. 2020; Bhanot, Rao, and Deshmukh 2016; Ingarao et al. 2012; Jayal et al. 2010; Labuschagne, Brent, and van Erck 2005; Linke et al. 2013; Lu et al. 2012; Mani et al. 2014; Olinto 2014; Saad, Nazzal, and Darras 2019; Singh et al. 2009). Several studies have introduced sustainability assessment indices such as the Product Sustainability Index (ProdSI) to evaluate the sustainability of manufactured goods without considering process variables (Hapuwatte et al. 2016).

However, limited research has been conducted on assessing sustainability at the process level, particularly for processes other than machining operations (Abbas et al. 2019; Bhanot, Rao, and Deshmukh 2016; Haq et al. 2021; Jamil et al. 2021; Khan et al. 2020; Linke et al. 2013; Singh et al. 2020). While studies have proposed frameworks and metrics-based assessment methods for machining operations, processes such as casting, joining, shaping, and forming have received less attention (Sivarupan et al. 2021; Wu et al. 2020). For example, research on welding processes has utilised sustainability assessment tools like weight space partitioning and life cycle assessment (LCA) to compare different welding methods and evaluate their sustainability performance (Sproesser et al. 2016).

Efficient energy and material use in sheet metal forming and die casting processes have also been studied, with researchers conducting sensitivity analyses and developing computer-aided tools for sustainability assessment (Ingarao et al. 2012; Singh et al. 2009).

Additionally, the evaluation of industrial technical devices in manufacturing processes has been recognised as crucial for enhancing sustainability. These devices offer advantages and opportunities but require critical evaluation to ensure effective sustainability assessment (Table 1).

Table 1. Advantages and disadvantages of industrial devices for sustainability in machining processes.

Sources	Industrial Devices	Advantages	Disadvantages
(He and Bai 2021; Jamwal et al. 2021; Verl et al. 2019)	Robotics and Automation	Improved productivity and reduced cycle times Enhanced workplace safety by minimising human involvement Consistency in quality and reduced rework	High initial costs of implementation Maintenance and repair requirements Limited flexibility for customised or small-scale production
(Abubakr et al. 2021; Imad et al. 2022; Kandavalli et al. 2023; Kuntoğlu et al. 2021; McAfee et al. 2022)	Advanced Sensing and Monitoring Systems	Real-time data collection for process optimisation Detection of anomalies and preventive maintenance Improved quality control and defect detection	Initial investment and installation costs Data management and analysis challenges Integration with existing systems and compatibility issues
(Nagalingam, Toh, and Yeo 2023; Nor et al. 2020; Xiao et al. 2019)	Programmable Logic Controllers (PLCs)	Enhanced process control for improved stability and quality Energy optimisation through efficient equipment usage Proactive maintenance and diagnostics for reduced downtime Flexibility in adapting to changing production requirements	Complexity in design and programming High initial costs and specialised expertise required Regular maintenance and software updates needed Dependence on electrical infrastructure
(Mawson and Hughes 2020; Ortega Del Rosario, Beermann, and Chen Austin 2023)	Building Management Systems (BMS)	Energy management and optimisation for reduced consumption Centralised control and monitoring of building systems Improved occupant comfort and indoor environmental quality Integration with renewable energy sources for sustainability	Complex installation and integration process Costs associated with equipment retrofitting Maintenance and software updates required Limited compatibility with legacy systems
(Ashima et al. 2022; Kang 2020; Rahmani et al. 2023)	Human-Machine Interface (HMI)	Intuitive and user-friendly interface for operators Real-time monitoring and control of machining processes Quick response to alarms and notifications for timely intervention	Potential for operator error if not properly trained Risk of information overload with excessive data display
(Al-Naggar et al. 2021; Lee et al. 2019; Li et al. 2019)	Predictive Maintenance Systems	Condition-based monitoring for early fault detection Reduced unplanned downtime and maintenance costs Optimised maintenance schedules for increased equipment lifespan	Initial investment and implementation costs Integration with existing equipment and data systems Reliance on accurate data and analytics for effective predictions
(Errigo, Choi, and Kissock 2022; Kumar et al. 2022)	Variable Frequency Drives (VFDs)	Energy savings through motor speed control and optimisation Reduced wear and tear on mechanical components Smooth starting and stopping of motors, minimising power surges Compatibility with energy management systems for optimisation	Higher initial costs compared to traditional motor starters Potential harmonics and electrical noise generation Additional complexity in installation and programming

2.2. Sustainability assessment frameworks

The concept of sustainability has gained increasing significance in modern work environments and culture, with manufacturers and consumers alike emphasising sustainable practices and products. However, there is still a lack of consensus among scholars and practitioners regarding the definition of sustainable development (SD) and sustainable manufacturing. The widely accepted definition of sustainable development is provided by the Brundtland commission (undefined.), while the Department of Commerce’s definition is widely recognised for sustainable manufacturing (Kim et al. 2015; Sangwan and Mittal 2015).

A literature review of sustainability assessment frameworks and initiatives in manufacturing organisations reveals a range of practices and tools employed by researchers. These include the balanced scorecard method, sustainable value stream mapping, multi-criteria decision making, indicator-based assessment, software and mathematical modelling, life cycle analysis, product-service systems, and sustainability indices (Chou, Chen, and Conley 2015; Hallstedt 2017; Herrmann et al. 2014; Rajak and Vinodh 2015; Salvado et al. 2015).

Various types of sustainability assessment tools have been identified, including financial, biophysical, and indicator tools, each with subcategories. It is recommended to use a well-balanced ‘indicator tool’ that combines financial and biophysical aspects and incorporates stakeholder value orientations. However, a well-organised sustainability assessment framework is still lacking, leading to the use of different evaluation methodologies (Gasparatos and Scolobig 2012).

Another study proposes a hierarchical approach to sustainability assessment in the manufacturing sector, drawing inspiration from Maslow’s hierarchy of needs. The framework focuses on addressing complex interactions between elements, performance measures, industry, society, and ecosystems. It incorporates a sustainability cone-based model and includes life cycle stages during the early design phase. The framework encompasses organisational spheres, stakeholder involvement, product and process management, and sustainability dimensions, providing a systematic and holistic evaluation of sustainability performance (Bhakar, Digalwar, and Sangwan 2018). The process flow of the framework is shown in Figure 1.

Figure 1. 5 level sustainability framework (Bhakar, Digalwar, and Sangwan 2018).

Hierarchical sustainability assessment framework for manufacturing. Resembles Maslow’s hierarchy with a bottom to top-shaped model. Five assessment levels: organisation, stakeholders, product, process, and sustainability. Considers interactions, life cycle stages, and various indicators for holistic evaluation.

A proposed framework provides a systematic structure for evaluating sustainability performance in manufacturing processes. It incorporates qualitative and quantitative indicators with weighting methods, conducts quantification and normalisation, and utilises multi-criteria decision-making (MCDM) methods for aggregation. Sensitivity analysis is employed to assess aggregated scores. While the framework offers a structured approach for decision-making, it has limitations. It lacks optimisation algorithms, restricting exploration beyond available datasets, and does not handle complex environments. Integrating optimisation algorithms could enhance its capability, enabling trade-off exploration and handling diverse datasets. Future research should address these limitations to achieve more robust sustainability assessment in manufacturing processes (Saad, Nazzal, and Darras 2019). The detailed framework can be seen in Figure 2.

Figure 2. MCDM based sustainability framework for manufacturing sector (Saad, Nazzal, and Darras 2019).

Illustration of a systematic framework for evaluating sustainability performance in manufacturing processes. It utilises qualitative and quantitative indicators, weighting methods, and multi-criteria decision-making (MCDM) for aggregation. Sensitivity analysis is employed.

Sustainability in the manufacturing sector requires a multi-dimensional approach, as highlighted in the literature (Gunasekaran and Spalanzani 2012). Several studies have presented hierarchical frameworks for achieving sustainability in manufacturing, including five-level hierarchical frameworks (Batterham 2006; Mittal and Sangwan 2014). The environmental, social, and economic impacts of manufacturing processes have led to an increased emphasis on sustainability. Various frameworks have been developed to measure sustainability in manufacturing, but most lack automated selection of the appropriate solution after global optimisation. To address this gap, this article presents a comprehensive framework consisting of seven levels: identification of sustainability dimensions and indicators, design of experiments, data extraction, data refinement, data modelling and optimisation, decision-making, and result validation. The proposed framework covers the full scope of sustainable manufacturing, considers the global environment, and sets absolute targets for sustainability improvement. Seventeen sustainable development goals are incorporated into the framework, emphasising the importance of people in sustainable development (Transforming Our World: The 2030Agenda for Sustainable Development .:. Sustainable Development Knowledge Platform n.d..). The hierarchical approach is adopted to handle complex interactions between elements, performance measures, industry, society, and ecosystems (Batterham 2006).

2.3. Machine parameter optimisation for sustainability improvement

Rising manufacturing energy use has contributed to the energy problem and global warming, with manufacturing responsible for a significant portion (36%) of carbon dioxide emissions and energy consumption (Cai et al. 2019; Spiering et al. 2015; undefined). To address these challenges, optimising manufacturing processes, particularly in machine tools and turning operations, is crucial for reducing energy consumption and improving environmental sustainability (Givehchi, Haghighi, and Wang 2017; Hu et al. 2015).

Optimisation techniques play a vital role in enhancing sustainability in manufacturing by reducing energy consumption, waste generation, carbon footprint, and raw material usage while improving product quality (Eheliyagoda et al. 2022; Obaideen et al. 2022; Pérez-Fernández, Sebastián, and González-Gaya 2022; Pimenov et al. 2022; Roslan et al. 2022). Green machining techniques, such as dry machining, cryogenic machining, and minimum quantity lubrication, further contribute to sustainability by minimising environmental impact (Memić, Hasković Džubur, and Avdagić-Golub 2022).

By employing optimisation techniques, manufacturing firms can achieve greater process efficiency, cost reduction, and improved sustainability, aligning with the goals of energy efficiency and environmental responsibility (Singh et al. 2020; Vats et al. 2022). These techniques help identify optimal machining parameters that optimise material removal rate, surface quality, and energy consumption, leading to sustainable manufacturing practices

Table 2 provides an overview of commonly studied machining processes, their parameters, and responses for optimisation. However, few studies have connected these responses to sustainability dimensions, hindering the quantification of sustainability in machining. To address this, our study aims to develop a framework that links sustainability with optimisation responses in machining. The proposed framework consists of seven levels and aims to calculate a numerical sustainability score. This framework can be applied to any machining process, addressing the need for a systematic approach to assess sustainability performance.

Table 2. Machining parameter for optimisation.

Source	Machining Type	Machining Parameters	Measured Responses
(Daniyan et al. 2019; Feng et al. 2022; Kosarac et al. 2022; Lee et al. 2020; Senthilkumar et al. 2021; Sharma et al. 2022; Wu and Lin 2021; Yan et al. 2021)	Milling	Cutting depth (radial and axial), cutting feed, cutting speed and machining condition i.e. dry, MQL or Cryogenic	Surface Roughness, Tool Flank Wear, Energy Consumption, Material Removal Rate, Cutting force, Profit, Moment of force, Machining time,
(Abas et al. 2020; Alshibi, Nasreldin, and Pervaiz 2023; Asgari, Delavar, and Sedighi 2023; Butola et al. 2020; Chate et al. 2021; Kumar Parida and Maity 2019; Lakshmana; Kumar et al. 2022; Marakini et al. 2022; Pathapalli et al. 2019; Pawanr, Kant Garg, and Routroy 2019; Radha Krishnan and Ramesh 2020; Zhou et al. 2019)	Turning	Cutting depth (radial and axial), cutting feed, cutting speed and machining condition i.e. dry, MQL or Cryogenic	Cutting force, Material Removal Rate, Tool Wear, Surface Roughness, Cutting Temperature, Micro-hardness, Chip Compression Ratio, Chip Segmentation Ratio, Saw-tooth Pitch, Land Width, Chip Contact Length, Shear Angle, Microstructure Changes
(Camposeco-Negrete 2021; Kam, İ̇pekçi, and Argun 2022; Kumar Debta, Mishra, and Masanta 2020; Payal et al. 2023; Ram Prasad et al. 2019; Shandilya, Rouniyar, and Saikiran 2020; Sharma, Vates, and Bansal 2021; Somasundaram, Pradeep Kumar, and Mathan 2020)	Electronic Discharge Machining (EDM)	Pulse on/off timings, electrode discharge current, Tool material, Angle of Cutting, wire tension,	Surface roughness, Electrode Wear Ratio, Material Removal Rate, machining time, Surface Roughness, Hole Overcut, Circularity at Entrance, Circularity at Exit, Cutting Velocity, Kerf width, Taperness

3. Proposed sustainability assessment framework

Sustainability has become an essential aspect of modern manufacturing. To assess the sustainability of a manufacturing process, it is necessary to have a robust and comprehensive framework. This article proposes a sustainability assessment framework that consists of seven levels, as discussed earlier, and an additional feature of best optimal solution selection among various. The proposed framework is designed to help numerically calculate the sustainability of any well-defined process and select the best optimal solution. Figures 3 and Figure 4 below shows a graphical representation of the proposed sustainability assessment framework. Moreover, complete process flowchart of the framework can be seen in Figures 5a, 5b.

Figure 3. Proposed framework.

Illustration of the list of stages of proposed framework: Indicator ID, Experiment Design, Data Extraction, Data Refinement, Data Modelling, Decision Making, Result Validation.

Figure 4. Technical representation of proposed framework.

Illustration depicting the technical approach to machine parameter optimisation. The process begins with the selection of indicators, followed by the design of experiments. Data extraction is carried out during machining, with the data stored over the cloud. The collected data is then refined and subjected to sensitivity analysis. Process modelling and optimisation are performed, leading to decision making. Sustainability assessment and validation are conducted to ensure the effectiveness of the optimisation process.

• Identification of indicators (Independent and Dependent Variables):

Figure 5. Process flowchart of proposed framework.

Figure 5. (Continued).

Read the detailed description of this figure

The illustration showcases a streamlined machine parameter optimization process. It initiates with identifying machine parameters and responses, categorized as measured or calculated. Measured responses involve sensor identification, while calculated responses establish parametric relationships. Subsequent steps include designing experiments, extracting and refining data to determine dominance. If dominance is observed, it proceeds to the modeling stage, adjusting hyperparameters for accuracy. The optimization block employs multi-objective optimization algorithms to generate an optimal solution set. In the decision-making stage, Multi-Criteria Decision Making (MCDM) techniques select the best solution validating sustainability assessment. This comprehensive process ensures efficient machine parameter optimization through strategic data-driven steps, from initial identification to final decision-making.

Identify the key machining parameters that significantly affect the machining process of magnesium alloys. These parameters may include cutting speed, feed rate, depth of cut, tool material, tool geometry, and cooling/lubrication techniques (Independent variables). Determine the response variables that reflect the performance and quality of the machining process, such as surface roughness, material removal rate, cutting forces, tool wear, and dimensional accuracy (Dependent Variables). Section 3.1 provides a comprehensive discussion on the various indicators identified in the literature.

• Design of Experiment:

Utilise Taguchi orthogonal array to design a set of experiments that efficiently explore the effects of different machining parameters on the response variables. Assign specific levels or values to each parameter in the orthogonal array to cover a wide range of conditions while minimising the number of experimental runs. Section 3.2 elaborates on additional methodologies for experimental design that have been documented in the literature

• Data Extraction through Sensor Fusion:

Install suitable sensors and measurement devices to collect data during the machining experiments. Capture relevant process variables, such as temperatures, vibrations, forces, and displacements, using sensor fusion techniques that integrate data from multiple sensors. This fusion process ensures comprehensive and accurate data collection. elaborates on additional methodologies for experimental design that have been documented in the literature. Section 3.3 presents a comprehensive list of the key machining parameters that significantly impact the economic, environmental, and social sustainability of the machining process. Additionally, it discusses the specific measuring devices, such as sensors, that are necessary for monitoring and extracting data related to these parameters.

• Data Refinement and Dimension Reduction:

The collected data is refined and analysed to evaluate the effects of different machining parameters on the response variables. A main effect plot is constructed to visualise the impact of each parameter on the response variables and identify the significant factors influencing the machining process. In Section 3.4, the study delves into the intricate steps of the data refinement process. Specifically, it discusses the utilisation of S/N analysis (Signal-to-Noise analysis) and PCA (Principal Component Analysis) as techniques for dimensional reduction in machine parameter optimisation. The section provides a detailed exploration of how these methods are applied to refine and streamline the data, enabling a more focused and efficient optimisation process.

• Data Modelling:

Develop mathematical models or utilise machine learning techniques to establish the relationships between the machining parameters and response variables. Use regression analysis or advanced algorithms to create models for each response variable, such as surface roughness prediction, material removal rate estimation, or cutting force optimisation. Tune the efficiency of each model through hyperparameter optimisation to enhance accuracy and predictive capability. Section 3.5 delves into the various data modelling techniques found in literature, employed to develop an efficient model that accurately represents the machining process. These techniques aim to capture the complex relationships between the machining parameters and the corresponding response variables, enabling the creation of a robust and reliable model.

• Multi-objective Optimisation of models:

Apply multi-objective optimisation techniques to optimise the machining process for magnesium alloys. Define objective functions considering sustainability and performance aspects, such as minimising energy consumption, reducing tool wear, and maximising material removal rate or surface quality. Adjust hyperparameters of the optimisation process to obtain optimal solutions that strike a balance between conflicting objectives. Section 3.5 explores the diverse range of multi-objective optimisation algorithms utilised in the literature for optimising machining parameters.

• Decision Making (Ranking Optimal Solutions using MCDM Techniques):

Rank the obtained optimal solutions achieved through multi-objective optimisation and evaluate their sustainability values using Multiple-Criteria Decision Making (MCDM) techniques. Assess sustainability metrics by considering economic, environmental, and social factors associated with the machining process. Employ MCDM methods to compare and rank the solutions, ultimately selecting the most sustainable alternative. This approach ensures a comprehensive evaluation of the optimal solutions and facilitates informed decision-making based on the identified sustainability criteria. Section 3.6 provides an overview of various classic Multiple-Criteria Decision Making (MCDM) approaches found in the literature that have been applied to similar types of applications.

• Validation:

Validate the selected optimal solution by conducting a machining experiment based on the identified machining parameters. Implement the machining process under the optimised conditions and evaluate its performance by comparing the experimental results with the predicted values from the data models. Analyze any discrepancies or deviations, and make necessary adjustments to further optimise the process if required.

By following this specific methodology, the proposed framework aims to achieve sustainable machining of magnesium alloys, considering factors such as efficiency, performance, and sustainability in the overall process. Section 3.7 explores additional validation approaches employed in the literature for similar types of applications.

3.1. Identification of indicators

The proposed sustainability assessment framework begins with the identification of indicators, which are measures of environmental, social, and economic sustainability (Azapagic 2004). This involves conducting a literature review, consulting with stakeholders, and evaluating the indicators based on criteria such as relevance and reliability (Dubravská et al. 2020; Schaltegger, Bennett, and Burritt 2006; Vishnu, Varghese, and Gurumoorthy 2023; Welford 2016). The selected indicators should cover environmental, social, and economic aspects and can include factors like greenhouse gas emissions, employee health and safety, and profitability. These indicators are then used to assess the sustainability performance of the manufacturing process and track improvements over time. Table 3 presents a range of indicators used in various systems to assess sustainability, highlighting the importance of sustainable materials and manufacturing processes (ISO 14,040 guidelines, LCA methodology).

Table 3. Sustainability assessment systems.

Sustainability Rating Systems	No. of identified indicators
2006 Environment Performance Indicators (Yale Center For Environmental Law And Policy-YCELP-Yale University, Center For International Earth Science Information Network-CIESIN-Columbia University, & World Economic Forum-WEF 2006)	19
Development Indicators (Wong 2014)	50
Ford Product Sustainability Index (Schmidt 2006)	8
General Motor Metrics for Sustainable Manufacturing (General Motors Metrics for Sustainable Manufacturing—LABORATORY FOR SUSTAINABLE BUSINESS n.d.)	46
Environmental Indicators for European Union (EEA indicators—European Environment Agency n.d.)	60
Global Report Initiative (GRI - Standards n.d..)	70

3.1.1. Environmental indicators

The environmental dimension of sustainability focuses on preserving natural resources, maintaining a balance between resource consumption and regeneration, and avoiding exceeding planetary boundaries (Bag, Panda, and Kumar Sahoo 2022; Gerten et al. 2019; Jie et al. 2023; Oláh et al. 2020). Environmental indicators serve as early warning systems and monitoring tools for tracking environmental damage. These indicators include metrics such as wastewater generation, water consumption, and material usage/recycling, as shown in Table 4 (Estoque et al. 2020).

Table 4. List of environmental indicators.

Sources	List of Environmental Indicators
(Cai et al. 2019; Mohsin et al. 2019; Pata 2021)	Carbon Emission
(Peter et al. 2021)	Wastewater produced
(Pinto 2020)	Water Consumption
(Bonilla Hernández et al. 2019)	Coolant Consumption
(Mahesh et al. 2021)	Cutting Temperature
(Kravchenko, Pigosso, and McAloone 2019)	Specific Energy Consumption

3.1.2. Economic indicators

The economic dimension of sustainability in manufacturing is concerned with the financial well-being of stakeholders and the impact on local and global economic systems (Sartal et al. 2020; Simpson and Tamayo 2020). It encompasses all economic transactions, including intangible assets and unconventional financial indicators (Roinioti and Koroneos 2019). However, defining what needs to be sustained from an economic perspective is not always clear (Lee et al. 2020). Compared to the environmental aspect, the economic dimension of sustainability has received less research attention, resulting in a lack of comprehensive indicator frameworks (Kim, Kim, and Lee 2013). Economic indicators used in manufacturing operations mainly focus on costs and profits, as shown in Table 5.

Table 5. List of economic indicators.

Sources	List of Economic Indicators
(Gupta et al. 2021)	Machining Cost
(Saxena et al. 2020)	Surface Quality
(Yadegaridehkordi et al. 2020)	Material Removal Rate
(Barni et al. 2022)	Profit

3.1.3. Social indicators

The social dimension of sustainability in manufacturing encompasses equity, accessibility, participation, and other elements that foster societal cohesiveness (Akcali and Cahantimur 2022). However, quantifying and incorporating the social dimension into assessments has been challenging, resulting in limited research and practice focus (Padilla-Rivera, Russo-Garrido, and Merveille 2020). Few established social indicators exist for the manufacturing industry, despite the importance of considering positive and negative social consequences (Nara et al. 2021). More attention is needed to assess intangible qualities such as consumer satisfaction (Morkunas and Rudienė 2020). Social indicators, both qualitative and quantitative, provide insights into the social dimension. Table 6 presents a selection of social indicators reported in the literature, capturing aspects like satisfaction with wages, work-related accidents, and employee turnover rates. To comprehensively evaluate social sustainability, it is essential to utilise a combination of qualitative and quantitative indicators (Mies and Gold 2021).

Table 6. List of social indicators.

Sources	List of Social Indicators
(Hegab, Darras, and Kishawy 2018)	Working environment noise level
	Working environment illumination level
	High-temperature surface exposure
	Atmospheric Conditions
	Mist/Dust Level

3.2. Design of experiment

Design of Experiments (DOE) is a statistical tool used to systematically analyse experiments and optimise machining processes. Various DOE methodologies are utilised, including full factorial design, fractional factorial design, response surface methodology (RSM), and the Taguchi method. The selection of the DOE methodology depends on factors such as problem complexity, the number of input variables, and available resources. Full factorial design offers a comprehensive analysis but can be time-consuming and costly (Kechagias et al. 2020). Fractional factorial design reduces the number of experiments while providing valuable insights (Horvath, Matyasi, and Dregelyi-Kiss 2014). RSM helps identify optimal input combinations for desired outcomes (Cardoso et al. 2021). The Taguchi method focuses on robust parameter settings, reducing variability and noise in the process (Yaka 2021). The choice of DOE methodology should consider the specific requirements and constraints of the machining problem (Daniel et al. 2020).

3.3. Data extraction

Data extraction in machining involves collecting and transforming raw data from various sensors to optimise machining parameters and improve efficiency. Sensor fusion combines data from multiple sensors, such as cutting force (Shah et al. 2022; Zarrouk et al. 2022), vibration (Kouahla et al. 2022; Sahoo et al. 2023), acoustic emission (Dixit, Sharma, and Kumar 2022), temperature (Delijaicov et al. 2023; Soori and Arezoo 2023), and power consumption (Li et al. 2023), to provide a comprehensive view of the machining process. This data can be used to analyse tool wear (Cheng et al. 2023; Nguyen et al. 2022), surface quality (Ishfaq, Rehman, and Wang 2023; Xie et al. 2023), chip formation (Bhuiyan, Choudhury, and Dahari 2014), and other factors.

3.4. Data refinement

Data refinement is a critical step in machine parameter optimisation, involving techniques such as data cleaning (Diao, Yan, and Gao 2022), transformation (Aydemir, Taskaya Temizel, and Temizel 2022), aggregation (Department of Electrical Engineering, Jamia Millia Islamia, New Delhi, India, 110001, Singh, J., din, Z., & Department of Electrical Engineering, Jamia Millia Islamia, New Delhi, India, 110001 2022), sampling (Huang 2022), and feature selection (El-Kenawy et al. 2023). Data cleaning corrects errors, while transformation standardises or normalises data. Aggregation summarises data, sampling selects subsets for analysis, and feature selection focuses on relevant variables. The most commonly used algorithms for feature selection are Principal Component Analysis (PCA) and Single-to-Noise ratio Analysis (S/N). PCA reduces dimensionality, capturing variance in the data, while S/N ratio identifies influential factors. PCA combines original features, and S/N ratio measures signal strength relative to noise. These techniques enhance the accuracy and reliability of data in machine parameter optimisation.

3.5. Data modelling and optimisation

Data modelling is the process of creating a mathematical or statistical model to represent a real-world system or process. Optimisation, on the other hand, involves finding the best set of input parameters to maximise or minimise a particular output variable based on the model (Qiao et al. 2022). In the context of machine parameter optimisation, data modelling and optimisation are used to predict and optimise machining parameters, such as cutting speed, feed rate, and depth of cut, to improve machining efficiency and reduce manufacturing costs. Data modelling and optimisation are important steps in machine parameter optimisation that involve creating a mathematical model of the system and using optimisation algorithms to improve its performance. These steps are often used one after the other to ensure that the model accurately represents the system and that it is optimised to achieve the desired response.

After data refinement and feature selection, the refined data is used to create a mathematical model of the system. Each response is mapped to the most influential machining parameters obtained after S/N or PCA analysis. This helps the model increase its efficiency and avoid disturbances. Once an accurate model is generated, it can be optimised using one of the many algorithms available to obtain the minimum or maximum values of the responses. Table 7 represents the literature discussing all the algorithms currently used in the field of machine parameter optimisation.

Table 7. Review of modelling and optimisation in machine parameter optimisation.

Sources	Methodology
(Kosarac et al. 2022)	Design of experiments using the Taguchi method, training ANN with 4-10-1 network structure using the BR algorithm for training.
(Kam, İ̇pekçi, and Argun 2022)	Design of experiments and optimisation using Taguchi method
(Paturi et al. 2022)	Full factorial design, RSM, and desirability function approach
(Kittali et al. 2022)	Taguchi technique is used for optimising the machining parameters.
(Feng et al. 2022)	Design of experiments using the Taguchi method, MOGA for optimisation.
(Lakshmana Kumar et al. 2022)	TOPSIS method for optimisation, Design of experiments using the Taguchi method, ANOVA for analysis of parameter and responses
(Mercy et al. 2022)	Regression modelling to establish a mathematical connection between the input parameters and the output responses, a GA for optimisation, and a surface morphology analysis for validation. The design of experiments was made utilising the Taguchi technique.
(Yan et al. 2021)	The milling process was simulated using the FEM, the Johnson-Cook constitutive model, and the RSM for assessing and optimising the milling settings.
(Payal et al. 2023)	Design of experiments and optimisation using Taguchi method and ANOVA for analysis of parameter and responses
(Abas et al. 2020)	Taguchi method, composite desirability function, and CRITIC for optimisation are used in the design of experiments. SEM and EDX were used for validation. ANOVA for parameter and response analysis.
(Lee et al. 2020)	Nelder-Mead simplex method for optimisation
(Pathapalli et al. 2019)	Box-Behnken experiments were used to compare multi-objective optimisation utilising WASPAS versus MOORA
(Maneiah et al. 2020)	The AWJM process parameters were optimised using the Taguchi-based ANOVA method. The significance of process parameters was assessed using the L9 orthogonal array.
(Sanjeevi, Kumar, and Krishnan 2021)	RSM with ANOVA and regression analysis to predict optimal solution
(Daniyan et al. 2019)	RSM
(Pawanr, Kant Garg, and Routroy 2019)	TOPSIS and GRA for ranking and optimisation
(Wu and Lin 2021)	Elman neural network to predict with PSO
(Senthilkumar et al. 2021)	Design of experiments and optimisation using Taguchi method, MOGA
(Shandilya, Rouniyar, and Saikiran 2020)	Teaching–learning-based optimisation, GRA and GA for optimisation
(Butola et al. 2020)	Design of experiment using Taguchi technique, RSM and GA for optimisation
(Radha Krishnan and Ramesh 2020)	Design of experiments and optimisation using Taguchi method
(Sharma et al. 2022)	Design of experiment using Taguchi technique, Taguchi-grey approach and TLBO
(Sharma, Vates, and Bansal 2021)	Integrated route of fuzzy and GRA approach with Taguchi’s technique
(Ram Prasad et al. 2019)	MADM using AHP and TOPSIS for optimisation
(Kumar Parida and Maity 2019)	CCD and RSM with ANOVA
(Zhou et al. 2019)	c-PBOM-W and c-PBOM-P tables were created to evaluate carbon emissions. A multi-objective optimisation model was established, and an improved algorithm, NG-NSGA-II, was proposed to solve the model.
(Chate et al. 2021)	Using the Taguchi technique for experiment design, sustainability optimisation models MOORA and DEAR were used.
(Camposeco-Negrete 2021)	Orthogonal array, signal-to-noise ratio, means graphs ANOVA, and NSGA-II for multi-objective optimisation

3.6. Decision making and sustainability assessment

Machine parameter optimisation is a complex problem that involves multiple objectives, and the trade-offs between these objectives lead to multiple sets of optimal solutions. The selection of the most suitable solution among the obtained ones is done manually based on the advice of an expert in the field. However, this issue can be resolved by applying Multi-Criteria Decision Making (MCDM) techniques like TOPSIS (Technique for Order Preference by Similarity to Ideal Solution), GRA (Grey Relational Analysis), and many others (Bhuvanesh Kumar et al. 2023; Pinzon Amorocho and Hartmann 2022).

MCDM helps identify the most suitable solution based on applied weights and criteria. The weights can be applied using various algorithms available in the literature like Analytic Hierarchy Process (AHP) (Singh Sivam, Rajendran, and Harshavardhana 2022), Fuzzy weights (Ye et al. 2022), and others. The weight applied to each criterion represents its importance.

Moreover, the result obtained from MCDM can be validated using a sustainability assessment model that considers the economic, social, and environmental aspects of the optimised responses/indicators. The performance of each indicator can be calculated using the Equations 1–3. The graphical representation can be seen in Figure 6.

(1) $EconomicSustainabilityAssessment\left(ESA\right)=\sum \left(Weightofeconomicindicator\ast normalizedvalueofindicator\right)$(1)

(2) $EnvironmentSustainabilityAssessment\left(EnSA\right)=\sum \left(Weightofenvironmentindicator\ast normalizedvalueofindicator\right)$(2)

(3) $SocialSustainabilityAssessment\left(SSA\right)=\sum \left(Weightofsocialindicator\ast normalizedvalueofindicator\right)$(3)

Figure 6. Technical indicators to assess sustainability.

Illustration presenting a list of technical indicators for evaluating the economic, environmental, and social sustainability of manufacturing processes. The indicators encompass various aspects such as product quality, energy efficiency, waste generation, carbon emissions, worker safety, labour conditions, and economic performance. The comprehensive set of indicators provides a holistic approach to assessing sustainability across multiple dimensions of manufacturing operations.

3.7. Results validation

To validate the results obtained through parameter optimisation and sustainability assessment, a robust validation process is crucial. This process involves two key aspects: utilising sustainability assessment criteria (Kalateh, Kianifar, and Sardarabadi 2022) and conducting experimental validation (Mahfoud et al. 2022; Zamfirache et al. 2022).

Firstly, the framework emphasises the use of sustainability assessment criteria to evaluate the optimised parameters. These criteria encompass various dimensions, including environmental impact, social responsibility, and economic feasibility. By considering these criteria, the framework ensures that the optimised parameters not only improve the performance of the machining process but also contribute positively to the broader sustainability goals. The criteria can be tailored to specific industries or processes, allowing for a customised and comprehensive evaluation of sustainability.

Secondly, experimental validation is a critical step in the validation process. It involves conducting physical experiments or simulations to verify the performance of the optimised parameters in real-world or simulated scenarios. This validation helps to confirm that the optimised parameters indeed lead to improved machining process performance while aligning with sustainability objectives. By comparing the experimental results with the predicted outcomes from parameter optimisation, the framework ensures the reliability and applicability of the optimised parameters.

4. Case study

The case study presented in this section aims to implement the proposed framework for sustainability assessment of the machining process through machining parameter optimisation technique. The dataset utilised in this study comprises experimental data obtained from runs on a milling machine, conducted under diverse operating conditions. The current case is focused on determining the optimal machining parameters to machine the two materials found in the dataset i.e. cast iron and steel using the propose framework. Within this dataset, a total of 16 cases were examined Table 8, each comprising a varying number of runs. The number of runs conducted for each case was contingent upon the degree of flank wear observed between runs. These measurements were taken at irregular intervals up to a predefined wear limit and occasionally extended beyond it.

Table 8. Experimental conditions.

Case	Depth of Cut (DOC)	Feed	Material
1	1.5	0.5	1 – Cast iron
2	0.75	0.5	1 – Cast iron
3	0.75	0.25	1 – Cast iron
4	1.5	0.25	1 – Cast iron
5	1.5	0.5	2 - Steel
6	1.5	0.25	2 - Steel
7	0.75	0.25	2 – Steel
8	0.75	0.5	2 – Steel
9	1.5	0.5	1 – Cast iron
10	1.5	0.25	1 – Cast iron
11	0.75	0.25	1 – Cast iron
12	0.75	0.5	1 – Cast iron
13	0.75	0.25	2 – Steel
14	0.75	0.5	2 – Steel
15	1.5	0.25	2 – Steel
16	1.5	0.5	2 – Steel

By utilising this dataset, the case study aims to provide readers with a practical illustration of the proposed framework’s application, offering valuable insights into how the framework can be effectively implemented to optimise machining parameters while ensuring sustainability. Through this empirical investigation, readers will gain firsthand knowledge of the framework’s functionalities and its potential to enhance machining process performance and sustainability outcomes.

4.1. Identification of indicators

The dataset utilised in this study encompasses a range of response variables that serve as indicators for computing sustainability in the machining process. These response variables include Flankwear (VB), Machining time (t), Spindle motor AC current consumption (I), Spindle Vibration (V), and Spindle acoustic emission (A). Each of these variables plays a crucial role in assessing different aspects of sustainability.

The computation of economic sustainability involves reducing Flankwear (VB), Machining time (t), Spindle Vibration (V) and Spindle motor AC current consumption (I). These indicators directly impact the economic viability of the machining process by minimising tool wear, optimising the time required for machining, and reducing energy consumption.

Environmental sustainability, on the other hand, focuses on reducing Spindle motor AC current consumption (I). By minimising energy consumption, this indicator contributes to lowering the environmental impact associated with the machining process. The emphasis on energy efficiency aligns with the broader goal of sustainable manufacturing practices.

Additionally, social sustainability is evaluated through the indicator of Spindle acoustic emission (A). This variable allows for the assessment of the impact of machining on worker safety and well-being. By considering acoustic emissions, potential risks and disturbances to the workforce can be identified and mitigated, promoting a socially responsible machining environment.

4.2. Design of experiment

This study employed a factorial design L16 array and conducted 166 experiments, each replicated 16 times, to investigate the impact of machining parameters on sustainability indicators. The factorial design facilitated the simultaneous examination of multiple factors, while replication enhanced data precision and allowed for the evaluation of experimental variability. The sustainability indicators measured included Flankwear (VB), Machining time (t), Spindle motor AC current consumption (I), Spindle Vibration (V), and Spindle acoustic emission (A), representing economic, environmental, and social aspects. The comprehensive dataset obtained from the experiments provided a robust foundation for analysing the relationships between machining parameters and sustainability.

4.3. Data extraction

This study employed specific sensors to extract data during the machining process. The WD 925 acoustic sensor captured acoustic emission, while the 720,150 accelerometer measured vibrations. Additionally, the CTA 213 current sensor monitored spindle motor current. These sensors provided valuable insights into the behaviour, dynamics, and energy consumption of the machining operation. The data extracted from these sensors facilitated a comprehensive analysis of performance, efficiency, and sustainability indicators. By integrating these sensors and data extraction techniques, the study aimed to enhance understanding and optimise machining practices.

4.4. Data refinement

During the data refinement process, a main effect plot was generated to analyse the influence of each machining parameter on the material and responses as shown in Figures 7 and 8. Surprisingly, the main effect plot indicated that every response was significantly affected by every parameter. This implies that there was no specific parameter dominating any particular response. As a result, the reduction in dimension for modelling purposes was deemed not applicable for any of the responses. This finding highlights the complex interdependencies and intricate relationships between the machining parameters and the responses, suggesting that a comprehensive approach considering all parameters may be necessary for optimising the machining process.

Figure 7. Main-effect plot of responses (cast-iron).

Figure 8. Main-effect plot of responses (steel).

4.5. Data modelling

After the data refinement process and determining the necessary dimensions for modelling each response, the next step involves utilising linear regression models to obtain mathematical representations of the responses. These regression models not only consider the individual factors but also incorporate their interactions.

In the table below, the models for each material and their corresponding responses are presented. These models provide a quantitative understanding of the relationship between the machining parameters and the desired outcomes, enabling better control and optimisation of the machining process.

By employing regression analysis, it becomes possible to identify the significant factors and their impact on the responses, facilitating decision-making and process improvements. The accuracy of the prediction serves as an indicator of the reliability and effectiveness of the developed models. The obtained mathematical models serve as valuable tools for further analysis, optimisation, and decision support in the context of machining processes.

4.6. Data optimisation

In order to improve the performance and efficiency of the machining process, the generated data presented in Table 9 is subjected to an unconstrained optimisation using a classical genetic algorithm. The MATLAB Optimisation Toolbox was utilised to optimise the generated model in Table 9. Unconstrained optimisation techniques were employed, leveraging the flexibility provided by the toolbox. Hyperparameters, such as 500 generations and a population size of 50, were carefully tuned to obtain the best optimal solutions.

Table 9. Response regression models for both materials.

Cast Iron

Vb = 0.242 + 0.017 DOC + 0.033 feed + 0.047 DOC*feed

T = 85.7–41.9 DOC − 81.2 feed + 55.7 DOC*feed

I = 1.311 + 0.157 DOC + 0.273 Feed − 0.640 DOC*Feed

V = 0.3015–0.0159 DOC − 0.1135 feed + 0.1081 DOC*feed

A = 0.1565–0.0052 DOC − 0.0780 feed + 0.0152 DOC*feed

Steel

Vb = 0.823–0.318 DOC − 0.76 feed + 0.55 DOC*feed

T = 42.4–21.0 DOC − 58.3 feed + 32.1 DOC*feed

I = 1.3260 + 0.0494 DOC + 0.1053 feed − 0.1321 DOC*feed

V = 0.2980–0.0277 DOC − 0.0297 feed + 0.0705 DOC*feed

A = 0.1282–0.0166 DOC − 0.0098 feed + 0.0228 DOC*feed

The optimisation process aimed to enhance the performance and efficiency of the machining process by exploring the parameter space. The results of the optimisation are presented in the table below in Table 10 and 11, showcasing the optimal solutions achieved through the MATLAB Optimisation Toolbox for cast iron and steel respectively.

Table 10. Optimal set of solutions for cast iron material.

DOC	Feed	Cast Iron
		Vb	T	I	V	A
0.8	0.3	0.27678	63.628	1.3649	0.280674	0.1326
0.8	0.5	0.2909	78.58	1.3171	0.27527	0.1194
1.5	0.3	0.29855	115.76	1.3404	0.292245	0.1321
1.4	0.3	0.29544	88.738	1.3439	0.290592	0.1322
1.1	0.5	0.30305	60.5	1.2682	0.286715	0.1201
1	0.3	0.283	103.19	1.3579	0.28398	0.1325
1.5	0.5	0.31925	57.745	1.203	0.301975	0.1211
0.9	0.3	0.27989	88.979	1.3614	0.282327	0.1325
1.3	0.4	0.30174	107.625	1.2915	0.291642	0.1264
1.5	0.4	0.3089	57.51	1.2717	0.29711	0.1266
0.8	0.4	0.28384	64.42	1.341	0.277972	0.1260
1	0.4	0.291	67.22	1.3212	0.28344	0.1262
1	0.5	0.299	53.53	1.2845	0.2829	0.1199
0.9	0.5	0.29495	52.687	1.3008	0.279085	0.1197
0.9	0.4	0.28742	85.872	1.3311	0.280706	0.1261
1.4	0.4	0.30532	80.618	1.2816	0.294376	0.1265
1.1	0.3	0.28611	15.47	1.3544	0.285633	0.1324

Table 11. Optimal set of solutions for steel material.

DOC	Feed	Steel
		Vb	T	I	V	A
0.8	0.3	0.4726	46.63	1.365406	0.28385	0.0910
1.5	0.3	0.3655	65.635	1.372245	0.279265	0.0842
1.5	0.5	0.3785	44.345	1.353675	0.294475	0.0714
1.3	0.3	0.3961	39.34	1.370291	0.280575	0.0861
1	0.3	0.442	52.06	1.36736	0.28254	0.0890
1.5	0.4	0.372	40.545	1.36296	0.28687	0.0778
1.1	0.3	0.4267	33.589	1.368337	0.281885	0.0881
0.9	0.3	0.4573	46.456	1.366383	0.283195	0.0900
1.4	0.3	0.3808	49.438	1.371268	0.27992	0.0851
1.2	0.3	0.4114	53.638	1.369314	0.28123	0.0871
1.4	0.5	0.3828	16.15	1.35534	0.29372	0.0719

4.7. Decision making and sustainability assessment

The novelty of this framework lies in its incorporation of a multi-criteria decision-making (MCDM) technique to validate and explain the optimal solutions obtained. In this case study, the Grey Relational Analysis (GRA) method is employed for decision-making purposes. The upcoming sections discuss the step-by-step procedure for applying GRA and assessing sustainability within the proposed framework.

To facilitate the decision-making process, the values of the responses are normalised using Equation 4(4) $x_{ij}=\frac{\mathrm{m}\mathrm{a}\mathrm{x}\left(y_{ij}\right)-y_{ij}}{max\left(y_{ij}\right)-min\left(y_{ij}\right)}\left(smaller-the-better\right)$(4) and Equation 5(5) $x_{ij}=\frac{y_{ij}-\mathrm{m}\mathrm{i}\mathrm{n}\left(y_{ij}\right)}{max\left(y_{ij}\right)-min\left(y_{ij}\right)}\left(larger-the-better\right)$(5) . These equations ensure that the responses are evaluated based on the ‘smaller-the-better’ or ‘larger-the-better’ characteristic. In this particular case study, all the responses, including flankwear, machining time, current consumption, vibrations, and acoustic emission, are expected to be smaller for better performance.

By normalising the dataset, the responses are transformed into a normalised dataset, which is presented below. This normalisation process enables a comprehensive analysis and comparison of the responses to support the MCDM technique’s application and subsequent sustainability assessment within the framework.

(4) $x_{ij}=\frac{\mathrm{m}\mathrm{a}\mathrm{x}\left(y_{ij}\right)-y_{ij}}{max\left(y_{ij}\right)-min\left(y_{ij}\right)}\left(smaller-the-better\right)$(4)

(5) $x_{ij}=\frac{y_{ij}-\mathrm{m}\mathrm{i}\mathrm{n}\left(y_{ij}\right)}{max\left(y_{ij}\right)-min\left(y_{ij}\right)}\left(larger-the-better\right)$(5)

The normalised values obtained from Table 12 are used for calculating the sustainability metrics. In this analysis, the weight of each response is assigned a value of 1, indicating equal importance. The sustainability metrics are calculated based on the following considerations:

• Economic Sustainability (ESA): Economic sustainability is evaluated by considering the variables Vb (flank wear), T (machining time), I (spindle motor AC current consumption), and V (spindle vibration). Reducing these parameters leads to cost reduction and improved product quality, thereby enhancing economic sustainability.

• Environmental Sustainability (EnSA): Environmental sustainability is assessed by focusing on the variable I (spindle motor AC current consumption). By reducing the current consumption, energy usage is minimised, resulting in lower greenhouse gas emissions and improved environmental sustainability.

• Social Sustainability (SSA): Social well-being is linked to the variable A (spindle acoustic emission), which represents noise levels. By reducing acoustic emissions and noise levels, the working environment becomes more conducive, benefiting the well-being and comfort of workers. This contributes to enhanced social sustainability.

Table 12. Normalised solutions.

Cast Iron					Steel
Vb	T	I	V	A	Vb	T	I	V	A
1.000	0.520	0.000	0.798	0.000	0.000	0.384	0.368	0.699	0.000
0.668	0.371	0.295	1.000	1.000	1.000	0.000	0.000	1.000	0.349
0.487	0.000	0.151	0.364	0.034	0.879	0.430	1.000	0.000	1.000
0.561	0.269	0.130	0.426	0.029	0.714	0.531	0.105	0.914	0.249
0.381	0.551	0.597	0.571	0.945	0.286	0.274	0.263	0.785	0.100
0.854	0.125	0.043	0.674	0.010	0.939	0.507	0.500	0.500	0.674
0.000	0.578	1.000	0.000	0.872	0.429	0.648	0.210	0.828	0.149
0.927	0.267	0.022	0.736	0.005	0.143	0.388	0.316	0.742	0.050
0.412	0.081	0.453	0.387	0.467	0.857	0.327	0.053	0.957	0.299
0.244	0.581	0.576	0.182	0.453	0.571	0.242	0.158	0.871	0.199
0.834	0.512	0.148	0.899	0.500	0.838	1.000	0.910	0.050	0.973
0.665	0.484	0.270	0.694	0.487
0.477	0.621	0.497	0.714	0.964
0.572	0.629	0.396	0.857	0.982
0.749	0.298	0.209	0.796	0.493
0.328	0.350	0.515	0.285	0.460
0.780	1.000	0.065	0.612	0.015

By considering these factors the values were calculated and shown in Table 13 and 14 for cast iron and steel respectively, the sustainability metrics provide a comprehensive evaluation of the machining process. The goal is to optimise the machining parameters in a way that ensures economic efficiency, environmental responsibility, and social well-being are taken into account.

Table 13. Sustainability assessment of cast iron through normalised responses.

Cast Iron
DOC	Feed	Vb	T	I	V	A	ESA	EnSA	SSA
0.8	0.3	1.000	0.520	0.000	0.798	0.000	2.32	0.00	0.00
0.8	0.5	0.668	0.371	0.295	1.000	1.000	2.33	0.30	1.00
1.5	0.3	0.487	0.000	0.151	0.364	0.034	1.00	0.15	0.03
1.4	0.3	0.561	0.269	0.130	0.426	0.029	1.39	0.13	0.03
1.1	0.5	0.381	0.551	0.597	0.571	0.945	2.10	0.60	0.95
1	0.3	0.854	0.125	0.043	0.674	0.010	1.70	0.04	0.01
1.5	0.5	0.000	0.578	1.000	0.000	0.872	1.58	1.00	0.87
0.9	0.3	0.927	0.267	0.022	0.736	0.005	1.95	0.02	0.00
1.3	0.4	0.412	0.081	0.453	0.387	0.467	1.33	0.45	0.47
1.5	0.4	0.244	0.581	0.576	0.182	0.453	1.58	0.58	0.45
0.8	0.4	0.834	0.512	0.148	0.899	0.500	2.39	0.15	0.50
1	0.4	0.665	0.484	0.270	0.694	0.487	2.11	0.27	0.49
1	0.5	0.477	0.621	0.497	0.714	0.964	2.31	0.50	0.96
0.9	0.5	0.572	0.629	0.396	0.857	0.982	2.45	0.40	0.98
0.9	0.4	0.749	0.298	0.209	0.796	0.493	2.05	0.21	0.49
1.4	0.4	0.328	0.350	0.515	0.285	0.460	1.48	0.51	0.46
1.1	0.3	0.780	1.000	0.065	0.612	0.015	2.46	0.06	0.01

Table 14. Sustainability assessment of steel through normalised responses.

Steel
DOC	Feed	Vb	T	I	V	A	ESA	EnSA	SSA
0.8	0.3	0.000	0.384	0.368	0.699	0.000	1.299	0.368	0.000
1.5	0.3	1.000	0.000	0.000	1.000	0.349	2.000	0.000	0.349
1.5	0.5	0.879	0.430	1.000	0.000	1.000	2.139	1.000	1.000
1.3	0.3	0.714	0.531	0.105	0.914	0.249	2.055	0.105	0.249
1	0.3	0.286	0.274	0.263	0.785	0.100	1.499	0.263	0.100
1.5	0.4	0.939	0.507	0.500	0.500	0.674	2.246	0.500	0.674
1.1	0.3	0.429	0.648	0.210	0.828	0.149	1.859	0.210	0.149
0.9	0.3	0.143	0.388	0.316	0.742	0.050	1.435	0.316	0.050
1.4	0.3	0.857	0.327	0.053	0.957	0.299	2.065	0.053	0.299
1.2	0.3	0.571	0.242	0.158	0.871	0.199	1.747	0.158	0.199
1.4	0.5	0.838	1.000	0.910	0.050	0.973	2.798	0.910	0.973

The economic, environmental, and social sustainability metrics obtained previously are further normalised using the larger-the-better characteristic. This normalisation process aims to facilitate a fair and consistent comparison among the different sustainability metrics. The normalised metrics are presented in the Table 15.

Table 15. Normalised sustainability metrics.

Cast Iron			Steel
ESA	EnSA	SSA	ESA	EnSA	SSA
0.904	0.000	0.000	0.000	0.368	0.000
0.915	0.295	1.000	0.467	0.000	0.349
0.000	0.151	0.034	0.560	1.000	1.000
0.263	0.130	0.029	0.504	0.105	0.249
0.755	0.597	0.945	0.134	0.263	0.100
0.477	0.043	0.010	0.632	0.500	0.674
0.396	1.000	0.872	0.373	0.210	0.149
0.652	0.022	0.005	0.090	0.316	0.050
0.227	0.453	0.467	0.511	0.053	0.299
0.398	0.576	0.453	0.299	0.158	0.199
0.955	0.148	0.500	1.000	0.910	0.973
0.763	0.270	0.487
0.898	0.497	0.964
0.998	0.396	0.982
0.722	0.209	0.493
0.326	0.515	0.460
1.000	0.065	0.015

The performance of each experiment for specific sustainability metrics is evaluated by comparing the processed data values obtained through data pre-processing. The experiment with a value closest to 1 is considered to have the best performance for that response. To facilitate comparison, a reference sequence (X0) is established with reference values for each metric. The deviation sequences in Table 16 indicate the proximity of the comparability sequences to the reference sequence, enabling the determination of each experiment’s relative performance for different responses.

Table 16. Deviation sequence.

Cast Iron			Steel
ESA	EnSA	SSA	ESA	EnSA	SSA
0.096	1.000	1.000	1.000	0.632	1.000
0.085	0.705	0.000	0.533	1.000	0.651
1.000	0.849	0.966	0.440	0.000	0.000
0.737	0.870	0.971	0.496	0.895	0.751
0.245	0.403	0.055	0.866	0.737	0.900
0.523	0.957	0.990	0.368	0.500	0.326
0.604	0.000	0.128	0.627	0.790	0.851
0.348	0.978	0.995	0.910	0.684	0.950
0.773	0.547	0.533	0.489	0.947	0.701
0.602	0.424	0.547	0.701	0.842	0.801
0.045	0.852	0.500	0.000	0.090	0.027
0.237	0.730	0.513
0.102	0.503	0.036
0.002	0.604	0.018
0.278	0.791	0.507
0.674	0.485	0.540
0.000	0.935	0.985

The grey relational coefficient is utilised to assess the proximity between xij and x0j, where a larger coefficient indicates a closer resemblance. The coefficient can be calculated using the equation 6, with $\xi$ representing the distinguishing coefficient ranging from 0 to 1. In this study, $\xi$ is assumed to be 0.5, reflecting a moderate level of distinguishability. A smaller value of zeta indicates a higher level of distinguishability between the compared values.

(6) $\gamma \left(x_{oj},x_{ij}\right)=\frac{\mathrm{\Delta }min+\xi \mathrm{\Delta }max}{{\mathrm{\Delta }}_{ij}+\xi \mathrm{\Delta }max}$(6)

The grey relational grade is a measure used to quantify the similarity between a comparability sequence and a reference sequence in the grey relational space. It is computed as a weighted sum of grey relational coefficients, representing the degree of closeness between the sequences. The grey relational grade is determined using the provided equation 7, where n represents the number of responses and wj denotes the weight assigned to each response.

(7) $r\left(X_{o,}{X}_i\right)=\sum _{j=a}^n{w}_j\gamma \left(x_{oj},x_{ij}\right)$(7)

There are various methodologies to calculate the weight of each response like analytical heretical process (AHP), Fuzzy weights, etc. In this study, the equal weight approach is opted and weight of every response is kept at 1. The grey relational grade serves as an indicator of the level of similarity between the comparability sequence and the reference sequence. The experiment with the highest grey relational grade indicates the best choice, as it exhibits the closest resemblance to the reference sequence.

Table 17 displays the grey relational coefficients and grades for each experiment. Notably, experiment number 14 (Cast Iron) and experiment number 11 steel, highlighted in yellow, demonstrates the best optimal combination of controllable parameters.

Table 17. Grey rational coefficients and grades.

Based on the Grey Relational Analysis (GRA), the optimal machining parameters for cast iron are determined to be a Depth of Cut (DOC) of 0.8 and a feed rate of 0.5. Similarly, for steel, the optimal parameters are a DOC of 1.4 and a feed rate of 0.5. These parameter combinations have been ranked as the top choices among all the optimal solutions obtained based on sustainability metrics.

4.8. Validation

In this case study, a framework for sustainability assessment of the machining process through parameter optimisation has been developed and applied to an available dataset. Due to the unavailability of specific testing conditions and environmental factors during data extraction, experimental validation was not conducted. Instead, the validation of the framework was performed using sustainability assessment metrics. The metrics considered economic, environmental, and social sustainability factors to evaluate the performance of the optimised machining parameters. The results obtained through the sustainability assessment metrics provide valuable insights into the sustainability of the optimised solutions. Although experimental validation was not possible, the rigorous analysis and evaluation of the dataset using sustainability metrics ensure the reliability and applicability of the framework. The findings contribute to the advancement of sustainable machining practices and serve as a reference for future research and applications.

4.9. Conclusion

In this case study, the objective was to determine the optimal machining conditions for cast iron and steel based on a provided dataset obtained through milling machining. The dataset consisted of responses such as flankwear, machining time, spindle current consumption, spindle vibration, and acoustic emission, which were linked to sustainability indicators for assessing the economic, environmental, and social sustainability of the machining process.

By following the proposed framework, a set of optimal solutions was obtained. These solutions were then analysed using the Grey Rational Analysis (GRA) which is a MCDM technique, and sustainability metrics were calculated to identify the best optimal solution. The findings revealed that for machining cast iron, a depth of cut (DOC) of 0.8 and a feed rate of 0.5 should be maintained, while for steel, a DOC of 1.4 and a feed rate of 0.5 are recommended. These machining parameters ensure sustainable machining for both materials, contributing to improvements in economic, environmental, and social aspects.

Through the integration of MCDM techniques and sustainability assessment metrics, this case study demonstrates the importance of considering multiple sustainability dimensions in optimising machining processes. The findings provide valuable insights for practitioners in the manufacturing industry to make informed decisions for sustainable machining operations. Further research and implementation of the proposed framework can contribute to the advancement of sustainable manufacturing practices and help mitigate the environmental impacts of machining processes while considering economic viability and social well-being.

5. Conclusion and future directions

This review paper highlights the prevalence of multi-objective optimisation techniques in machine parameter optimisation and sustainability assessment; it is evident that multi-objective optimisation techniques are widely used in this field. However, a key gap identified is the lack of a standardised procedure for selecting the best solution from the set of optimal solutions generated by these techniques. To address this gap, a holistic framework has been developed that incorporates multi-criteria decision-making techniques to evaluate the sustainability of machining processes. This framework enables an objective and comprehensive evaluation of the set of optimal solutions obtained through parameter optimisation.

In addition to the framework, a case study was conducted to validate its effectiveness. The case study focused on determining the optimal machining conditions for cast iron and steel using a provided dataset obtained through milling machining. By following the proposed framework, a set of optimal solutions was obtained. These solutions were then analysed using the Grey Rational Analysis (GRA) which is a MCDM technique, and sustainability metrics were calculated to identify the best optimal solution. The findings revealed that for machining cast iron, a depth of cut (DOC) of 0.8 and a feed rate of 0.5 should be maintained, while for steel, a DOC of 1.4 and a feed rate of 0.5 are recommended. These machining parameters ensure sustainable machining for both materials, contributing to improvements in economic, environmental, and social aspects.

Through the integration of MCDM techniques and sustainability assessment metrics, this case study demonstrates the importance of considering multiple sustainability dimensions in optimising machining processes. The findings provide valuable insights for practitioners in the manufacturing industry to make informed decisions for sustainable machining operations. Further research and implementation of the proposed framework can contribute to the advancement of sustainable manufacturing practices and help mitigate the environmental impacts of machining processes while considering economic viability and social well-being.

Looking ahead, several future directions can advance this field. Firstly, developing more sophisticated and efficient algorithms for multi-objective optimisation, capable of handling a large number of objectives and decision variables, will enhance optimisation outcomes. Additionally, integrating artificial intelligence and machine learning techniques into the framework can improve the accuracy and efficiency of the optimisation process.

Moreover, exploring the applicability of the framework across diverse machining processes and industries will yield valuable insights and expand practical implications. Investigating the economic aspects of sustainable machining, such as cost optimisation and resource efficiency, can provide further avenues for research. Furthermore, considering the social dimension of sustainability, including worker safety, fair labour practices, and community engagement, will contribute to a more holistic approach.

It is important to note that the manufacturing sector’s contributions to sustainable development extend beyond the scope of this research. While manufacturing drives economic growth, it must also embrace sustainability to mitigate environmental impacts and ensure the well-being of its workforce.

Abbreviation

Abbreviation	=	Details
AHP	=	Analytical Hierarchy Process
ANN	=	Artificial Neural Network
ANOVA	=	Analysis of Variance
AWJM	=	Abrasive water jet machining
CCD	=	Central Composite Design
CFRP	=	Carbon Fiber-Reinforced Plastic
CNC	=	Computer Numeric Control
CRITIC	=	Criteria Importance through Inter-Criteria
DEAR	=	Data Envelopment Analysis Ranking
DOE	=	Design of Experiment
EDM	=	Electrical Discharge Machining
EDX	=	Energy Dispersive X-ray Analysis
FEM	=	Finite Element Analysis
FSW	=	Friction Stir Welding
GA	=	Genetic Algorithm
GRA	=	Grey Rational Analysis
GRI	=	Global Reporting Initiative
IEA	=	International Energy Agency
ISO	=	International Standards Organization
LCA	=	Life-Cycle Analysis
MADM	=	Multi-Attribute Decision Making
MCDM	=	Multi-Criteria Decision Making
MEC	=	Machining Energy Consumption
MOGA	=	Multi-Objective Genetic Algorithm
MOORA	=	Multi-Objective Optimisation by Ratio Analysis
MOPSO	=	Multi-objective Particle Swarm Optimisation
MQL	=	Minimum Quantity Lubrication
NSGA	=	Non-dominated Sorting Genetic Algorithm
PBOM	=	Production Bills Of Material
PCA	=	Principal Component Analysis
PID	=	Proportional-Integral-Derivative
ProdSI	=	Product Sustainability Index
RSM	=	Response Surface Methodology
SDG	=	Sustainable Development Goals
SEM	=	Scanning Electron Microscope
TLBO	=	Teaching learning-based optimisation
TOPSIS	=	Technique for Order Performance by Similarity to Ideal Solution
WASPAS	=	Weighted Aggregated Sum Product Assessment
WBGT	=	Wet Bulb Globe Temperature
WEDM	=	Wire Electrical Discharge Machining
WEF	=	World Economic Forum

Acknowledgments

The dataset used in this research paper is the “Milling Data Set” obtained from the BEST lab at UC Berkeley, as acknowledged by A. Agogino and K. Goebel in 2007. This dataset is available in the NASA Ames Prognostics Data Repository, maintained by NASA Ames Research Center in Moffett Field, CA. (Source: http://ti.arc.nasa.gov/project/prognostic-data-repository).

Disclosure statement

‘This work was supported by the Government of Pakistan Planning Commission under Grant Establishment of National Center in Big Data & Cloud Computing (NCBC). The funding provided by the Government of Pakistan Planning Commission was instrumental in enabling the execution of this research project’.

Additional information

Funding

The Government of Pakistan Planning Commission under the Grant Establishment of National Center in Big Data & Cloud Computing (NCBC) supported this work.

Notes on contributors

Muhammad Atif

Muhammad Atif is currently a PhD student in the field of Computer Science at SZABIST University. He also holds the position of Assistant Professor at the same institution. His research interests encompass a variety of topics including mechanical design, optimal machining, renewable energy, embedded systems, and automation.

Faraz Junejo

Faraz Junejo obtained his PhD (Mechatronics) degree from Loughborough University, UK. He is currently working as Professor in Mechatronics Engineering Department at SZABIST, Karachi. His research interests include Machine Vision, Renewable Energy, Condition monitoring, Robotics and Engineering Mechanics. He is an active member of Higher Education Commission’s National Curriculum review committee for both Mechatronics and Mechanical Engineering.

Imran Amin

Imran Amin is an Professor and Head of Department at SZABIST Institute in Karachi. He has served as the Head for Center of Renewable Energy and Research from 2009 to 2012.