P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... 571–579 OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS IN A SUPPLY CHAIN OPTIMIZIRANJE IZBIRE TRAJNOSTNIH MATERIALOV ZA KLIMATSKE NAPRAVE V DOBAVNI VERIGI P. Sivaraman * , S. Santhosh Mechanical Engineering, Sri Krishna College of Technology, Coimbatore, Tamil Nadu Prejem rokopisa - received: 2023-07-13; sprejem za objavo - accepted for publication: 2023-09-13 doi:10.17222/mit.2023.940 In order to cut their carbon footprint and promote environmental sustainability, the majority of businesses have now turned to- wards sustainable practises in their manufacturing processes and supply networks. The use of sustainable materials has drawn a lot of attention recently as a crucial step in accomplishing these goals. Choosing the material that is most suited for a product can be difficult, despite the fact that there are many sustainable materials available. This study uses machine learning–aran - dom forest algorithm and multi-criteria decision analysis (MCDA) to optimise the use of sustainable materials in supply-chain operations. The study uses machine learning algorithms to analyse data on different sustainable materials, their characteristics and their effects on the environment. The study also investigates how an optimised material selection affects the whole supply chain, including the production, packing and shipping operations. The research offers a complete strategy for reducing the envi- ronmental effect of industrial processes by combining approaches from material engineering, supply chain management and ma- chine learning. The novelty of this work resides in its integration of material engineering and machine learning strategies to en- hance the supply chain choice of sustainable materials. As a notable example, the study highlights the potential of mycelium as a sustainable material for air conditioner components. Mycelium’s unique properties, such as its biodegradability, lightweight nature and adaptability position it as a promising candidate, enhancing the environmental profile of air conditioners. By incor- porating mycelium-based components, manufacturers can significantly reduce carbon emissions, resource consumption and waste generation throughout a product’s lifecycle. This investigation underscores both the viability of mycelium and the broader significance of innovative material choices in reshaping industries towards a more sustainable future. Through such advances, this research not only contributes to the air conditioning sector but also establishes a paradigm for sustainable material adoption with far-reaching positive implications. Keywords: random forest algorithm, sustainable materials, multi-criteria decision analysis, supply chain Zaradi zmanj{evanja oglji~nega odtisa in promocije trajnostno vzdr`nega okolja se je glavnina podjetij usmerila k trajnostni praksi v svojih proizvodnih procesih in dobavnih verigah. Za dosego tega cilja se kot klju~ne komponente v uporabi pojavljajo novi trajnostni materiali. Izbira najbolj primernega trajnostnega materiala za dolo~en izdelek je lahko zelo te`ka, ~eprav je na razpolago mnogo le-teh. V pri~ujo~i {tudiji avtorja opisujeta uporabo strojnega u~enja in sicer algoritma naklju~nega gozda (RFA; angl.: Random Forest Algorithm) in analizo odlo~itve na osnovi ve~ kriterijev (MCDA; angl.: Multi Criteria Decision Analysis) za optimiziranje uporabe trajnostnih materialov v dobavni verigi. V {tudiji sta avtorja uporabila algoritme strojnega u~enja za analizo podatkov razli~nih trajnostnih materialov, njihovih lastnosti in njihovega vpliva na okolje. V {tudiji sta prav tako raziskovala kako izbrani optimizirani material vpliva na celotno dobavno verigo, ki je vklju~evala proizvodnjo, embaliranje, pakiranje in vrste transporta. Izdelana {tudija ponuja celovito strategijo zmanj{evanja vpliva industrijskih procesov na okolje s kombiniranjem pristopov materialnega in`eniringa, mened`menta (upravljanja) dobavne verige in strojnega u~enja. Ta raziskava je primer uporabe novega pristopa pri zdru`evanju strategij materialnega in`eniringa in strojnega u~enja za pove~anje uporabe trajnostnih materialov v dobavnih verigah. Kot tipi~en primer sta avtorja v {tudiji osvetlila potencial uporabe micelija kot trajnostnega materiala za komponente klimatskih naprav. Micelij (podgobje ali mre`asto razrasle nitke) ima enkratne lastnosti; kot so njegova biolo{ka samo-razgradnja, majhna gostota in prilagodljivost. Zato se ta naravni material uvr{~a med obetavne kandidate za uporabo kot okolju prijazen material v klimatskih napravah. Z vklju~itvijo komponent na osnovi micelija bi lahko proizvajalci ob~utno zmanj{ali oglji~ni odtis (emisije ogljikovega dioksida), porabo surovin in nastajanja odpadkov v ~asu njihove uporabe oziroma dobe trajanja. V tej raziskavi avtorja ne opozarjata samo na mo`nosti uporabe micelija temve~ tudi na {ir{i pomen izbire inovativnih materialov pri preoblikovanju in usmerjanju industrije k bolj trajnosno usmerjeni prihodnosti. S tak{no usmeritvijo izvedena raziskava ne prispeva samo k razvoju novih klimatskih naprav, temve~ je tudi zgled za ve~jo uporabo drugih trajnostnih materialov z daljnose`nimi pozitivnimi u~inki. Klju~ne besede: algoritm naklju~nega gozda, trajnostni materiali, analiza odlo~itve na osnovi ve~ kriterijev, nabavna veriga 1 INTRODUCTION Companies all over the world are increasingly fo- cused on sustainable development to lessen their envi- ronmental impact, enhance their environmental perfor- mance and encourage a wise use of natural resources. The selection of sustainable materials, which have little environmental effect throughout their life cycles, is a key component of sustainable development. Choosing the right material for a product may be difficult, especially when other issues are taken into account, such as price, availability, quality, supply chain and environmental ef- fect. This project uses environmental impact reduction and machine learning to optimise the use of sustainable ma- terials throughout a supply chain. In order to choose the Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 571 UDK 519-6 ISSN 1580-2949 Original scientific article/Izvirni znanstveni ~lanek MTAEC9, 57(6)571(2023) *Corresponding author's e-mail: luckysivaraman@gmail.com (Parthasarathi Sivaraman) most ecologically friendly materials, the study explores machine learning approaches that can analyse historical data of different materials, including their qualities and environmental effect. This study uses advanced optimisa- tion techniques like random forest algorithm and MCDA to find sustainable materials with low carbon, energy footprint, toxicity and water usage, as well as materials that are recyclable. This research looked at sustainable materials in a va- riety of scenarios. However, it was primarily concerned with environmental evaluations and did not provide sys- tematic insights into optimisation approaches for sustain- able material selection. The innovative aspect of this work is the integration of machine learning techniques with materials engineering, which allows the optimisa- tion of material sustainability in supply chains. A machine-learning random forest algorithm is used in this study because of its capacity to handle large vol- umes of data and deliver insights into complicated chal- lenges. The algorithm can detect trends in past data by analysing it. It can find patterns in the selection of sus- tainable materials, identify the best suited resources for a specific product, and optimise the material selection across the whole supply chain by analysing historical data. This study incorporates sustainability by incorporat- ing sustainable development principles into machine learning and optimisation methodologies. It does not only encourage sustainable growth, but also provides a cost-effective strategy for material selection by minimis- ing the environmental impact while preserving the prod- uct quality and supply-chain efficiency. The study applies the multi-criteria decision analysis (MCDA) optimisation approach, assessing potential sus- tainable material solutions based on numerous factors. The MCDA model allows for the selection of sustainable materials based on a mix of environmental, social and economic criteria across several industries. The question arises why we chose air conditioners. The reason behind the selection of air conditioners as the focal application for this research stems from their sub- stantial energy consumption and environmental impact. Air conditioners are widely used across the globe, partic- ularly in regions with hot climates, contributing signifi- cantly to electricity consumption and consequently greenhouse gas emissions. They are therefore a suitable target for sustainability advancements. Several factors make air conditioners a suitable choice for this study: Energy Consumption: Air conditioners are known for their energy-intensive operation, which often relies on fossil fuels for electricity generation. Optimizing the ma- terials used for air conditioner components can lead to a reduced energy consumption and lower carbon emis- sions. Environmental Impact: Air conditioners, due to their energy consumption and refrigerants used, contribute to the emission of greenhouse gases and other pollutants. By selecting sustainable materials and optimizing their usage, the overall environmental impact of these devices can be significantly diminished. Lifecycle Considerations: The lifecycle of air condi- tioners involves stages like manufacturing, usage and disposal. A sustainable material selection can influence each of these stages, from reducing the raw material ex- traction and waste generation to improving the energy ef- ficiency during operation. Global Relevance: Air conditioning is essential in various sectors, including residential, commercial and in- dustrial ones. As a result, improvements in their sustainability can have a broad-reaching impact across diverse industries. Technological Innovation: The integration of ma- chine learning and optimization techniques for the selec- tion of sustainable materials in air conditioner manufac- turing presents an innovative approach to addressing environmental concerns in a practical and impactful manner. In summary, the goal of this research is to optimise the sustainable material selection in supply chain opera- tions through the use of machine learning techniques, en- vironmental effect reduction, and optimisation ap- proaches such as MCDA. Companies may use the findings of this study to take a more responsible and sus- tainable approach to material selection while preserving the supply chain efficiency and product quality. This re- search attempts to decrease the environmental effect by optimising the use of sustainable materials in air condi- tioner manufacturing processes, using machine learning techniques. The study takes a comprehensive approach, analysing the data on various sustainable materials, their properties and environmental impacts, as well as training machine learning algorithms to select the optimal mate- rial based on the criteria like carbon and energy foot- print, toxicity, water usage and recyclability. This re- search integrates material engineering, supply chain management and machine learning approaches to create a comprehensive solution for an environmental impact reduction. The findings of this study have the potential to significantly improve sustainability practises, reduce car- bon footprints and provide a more responsible approach to the material selection for the air conditioner manufac- ture. Finally, this research helps to promote sustainable development practises in the global air conditioning in- dustry. 2 LITERATURE REVIEW Material selection is a critical aspect of the sustain- able supply chain management, and several studies were conducted in this area. A review of the literature con- ducted for this study identified several works that exam- ined the sustainable material selection in supply chains. However, there is limited research available on the inte- P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... 572 Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 gration of machine learning techniques with materials engineering and optimization approaches. There is a study on the sustainable material selection in a manufacturing company using MCDA. The authors determined that the environmental impact was the most important criterion for material selection, followed by manufacturing cost and social responsibility. Their study recommended the use of MCDA in the material selection process as it considers multiple criteria in decision-mak- ing. 1–3 A research evaluated the sustainable material selec- tion process in a furniture supply chain and proposed a decision-making model based on the analytic hierarchy process (AHP). Their study revealed that sustainability, cost and supply chain efficiency were the most critical factors, influencing the material selection. 4–8 A few other articles explored the use of big data ana- lytics and machine learning techniques in the closed-loop supply chain management. They suggested that these techniques can be utilized for predicting de- mand, optimizing inventory and supply chain network design. They highlighted the importance of collecting large amounts of data related to materials, suppliers and other relevant factors for an effective material selec- tion. 9–13 A few articles discuss supply chain performances, highlighting the importance of considering environmen- tal, social and economic criteria when selecting sustain- able materials. They recommended the use of tools such as life cycle assessment, risk assessment and supplier performance assessment in a material selection pro- cess. 14–16 Lastly, a few researchers explored the use of the multi-criteria decision analysis (MCDA) for the sustain- able material selection in various industries. They deter- mined that clustering algorithms, classification algo- rithms and optimization algorithms are the most appropriate machine learning techniques for a sustain- able material selection. They recommended that a mate- rial selection process should incorporate factors such as energy consumption, carbon footprint and recyclability to minimize the environmental impact. 17–19 This review highlights the importance of sustainable material selection in supply chains, incorporating envi- ronmental, social and economic criteria. The study aims to integrate machine learning and optimization tech- niques in a sustainable material selection process and ex- pands the existing research in this area. The study aims to make a significant contribution to the field of sustain- able supply chain management by developing a model for a sustainable material selection using machine learn- ing and optimization techniques. The literature review also indicates the gaps that need to be addressed. Limited Optimization Approaches: The existing re- search on sustainable material selection often lacks an in-depth exploration of optimization techniques. By in- corporating machine learning and MCDA into the pro- cess, this study bridges the gap between sustainable ma- terial attributes and their optimal application within sup- ply chains. Insufficient Integration: The integration of machine learning and materials engineering is an underexplored avenue. This research seeks to fill in this gap by creating a holistic framework that seamlessly incorporates materi- als science and data-driven decision-making. Real-World Application: While many studies empha- size theoretical aspects, this research strives to provide practical implementation strategies. By focusing on air conditioners as a specific application, the study aims to transform theoretical findings into tangible sustainability improvements in a prominent industry. Lack of Comprehensive Criteria: Previous works of- ten considered singular criteria for a material selection, neglecting the multifaceted nature of sustainability. By incorporating diverse factors like environmental, social and economic aspects, this study introduces a more com- prehensive evaluation model. Industry-Specific Exploration: The existing literature may lack focus on specific industries, potentially limit- ing the practical impact of findings. This research ad- dresses this limitation by concentrating on the air condi- tioning sector, offering tailored insights for a critical field. In summary, this research seeks to bridge the identi- fied gaps in the existing literature by introducing a com- prehensive approach that integrates machine learning techniques and sustainability principles for a material se- lection. By addressing these limitations, the study not only contributes to academic discourse but also provides actionable insights for the industries striving to make in- formed, responsible and sustainable material choices. 3 EXPERIMENTAL PART In recent years, there has been a substantial increase in the use of machine learning (ML) in supply chain ac- tivities. Machine learning algorithms can analyse mas- sive volumes of data, detect trends and provide insights into difficult situations, assisting supply chain deci- sion-makers. Machine learning approaches can analyse historical data linked to the qualities and environmental effect of materials, recommend the most suitable alterna- tives, and optimise material selection decisions in the context of sustainable material selection. The random forest method is a supervised learning approach used in machine learning for both classification and regression issues. It is based on ensemble learning, which mixes several classifiers to tackle complicated problems and increase model performance. The ap- proach employs numerous decision trees on the subsets of a dataset and takes the average to increase the dataset’s predicting accuracy. The ultimate outcome is predicted based on the majority vote of projections from each decision tree. By employing a larger number of P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 573 trees in the forest, random forest helps to avoid the prob- lem of overfitting. The approach assumes real values for the feature variable and low correlations between the predictions from different trees. The random forest method is preferred because it requires less training time, predicts the output with high accuracy, and retains accu- racy even when substantial amounts of data are absent. The technique works by mixing N decision trees to form a random forest and generating predictions for each tree formed in the first step. It entails randomly picking K data points from the training set, creating decision trees associated with the selected data points, and repeating the procedure to determine the number of decision trees to be created. The programme predicts a new dataset by locating the predictions of all decision trees, and then allocates the new data points to the category that receives the most votes. Overall, the random forest method is a powerful tool for optimising the performance of machine learning models. The random forest machine learning method may be used to optimise a sustainable material selection for air conditioners in a supply chain, decreasing their environmental effect in the long run. The programme can analyse a dataset of various materials used in the manu- facture of air conditioners and categorise them based on their sustainability and environmental effect. The algo- rithm can properly forecast which materials are the most sustainable and have the least environmental footprint by integrating various decision trees and taking the majority vote. This can assist us to make educated material selec- tion decisions and lower the total environmental effect of the air conditioning supply chain. Furthermore, the algo- rithm’s capacity to effectively handle big datasets while maintaining accuracy even with missing data makes it a significant tool for optimising sustainable material choices in a variety of sectors. This study’s approach is divided into two stages: Phase 1: Combination of data gathering and material se- lection sustainability criteria Data collection covers material properties, environ- mental impact, social responsibility and economic fac- tors. Identification of sustainable material alternatives is based on a review of the literature and expert opinions. We use the multi-criteria decision analysis (MCDA), and select the most sustainable material solutions by allocat- ing weights to different criteria. Phase 2: Machine learning techniques and the multi-cri- teria decision analysis A dataset including information on material attributes and their environmental impact is developed based on the material alternatives chosen. Using clustering, classi- fication and optimisation methods, the dataset is cleaned, pre-processed and analysed. Clustering algorithms group materials based on their qualities, whereas classification algorithms categorise materials based on specific criteria. Finally, optimisation algorithms choose the most eco- logically friendly material options. To validate the effi- cacy of the suggested technique, the results of the ma- chine learning analysis are compared with the traditional method of material selection using MCDA with the ran- dom forest algorithm. The following stages are included in the experimental process of this study: 1. Data Gathering: Data on the characteristics and en- vironmental effects of various sustainable material alter- natives are acquired from reliable sources such as scien- tific publications and patents. The information is gathered for a variety of parameters, such as carbon foot- print, energy consumption, toxicity, water usage and recyclability. 2. Sustainability Criteria: Based on the literature study and expert inputs, a set of criteria for the sustain- able material selection in supply chains was identified. These criteria include environmental impact, social re- sponsibility and economic factors. 3. Multi-Criteria Decision Analysis (MCDA): To pick sustainable material alternatives, specified criteria are included into the multi-criteria decision analysis model. The MCDA model weights each factor to priori- tise the selection of materials with lower environmental, social and economic effects. 4. Machine Learning Analysis: Machine learning techniques such as clustering algorithms, classification algorithms and optimisation algorithms are used to ana- lyse the selected sustainable material alternatives. The study is carried out on a dataset that was constructed us- ing the environmental effect and property data gathered in step 1. Finally, the proposed methodology and experimental procedure include data collection on material properties and environmental impact, identification of sustainability criteria, selection of sustainable materials using the multi-criteria decision analysis, analysis of selected ma- terials using machine learning techniques, and validation of the effectiveness of the machine learning approach by comparing it with the MCDA model. The study’s goal is to help decision-makers choose the most sustainable ma- terials for supply chains. Material Selection: There are several sustainable ma- terials that could be selected for this study, depending on the specific application and industry. Here are 8 sustain- able materials which are considered for this study based on the environmental impact, social responsibility and economic factors as shown in Table 1. Random forest algorithm As the name suggests, "random forest is a classifier that contains a number of decision trees on various sub- sets of a given dataset and takes the average to improve the predictive accuracy of that dataset." Instead of rely- ing on one decision tree, random forest takes the predic- tion from each tree and based on the majority votes of predictions, it predicts the final output. A greater number P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... 574 Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 575 Table 1: Material selection Material Environmental impact Social responsibility Economic factors Picture Bamboo Highly sustainable and renewable Production can provide income for local com- munities Cost-effective and durable for construction and furni- ture Recycled plastics Reduces waste and pol- lution Working conditions of waste collectors and recyclers may be chal- lenging and hazardous Generally lower cost than virgin plastics Hempcrete Low-carbon and energy efficient Helps create jobs in sustainable agriculture and manufacturing sec- tors More expensive than tradi- tional concrete Mycelium Biodegradable at the end of life cycle, grown using waste materials Potential to create jobs in sustainable agricul- ture and manufacturing sectors More expensive than tradi- tional plastic materials Cork Highly sustainable, har- vesting process does not require the tree to be cut down Helps create jobs in cork production and harvesting More expensive than tradi- tional wood and other flooring materials Recycled steel Reduces energy con- sumption and carbon emissions May provide job oppor- tunities for waste col- lectors and recycling industries Can be less expensive than virgin steel due to reduced manufacturing costs Straw bales Highly sustainable and low carbon footprint May provide new job opportunities for those with less traditional construction experience Can be cheaper due to a lower cost of raw material Natural fibre fab- rics Less impact on the en- vironment, absence of toxic chemicals Sustainable production may provide job oppor- tunities for farmers and textile workers More expensive compared to synthetic materials of trees in a forest leads to higher accuracy and prevents the problem of overfitting. The diagram shown in Figure 1 explains the working of the random forest algorithm. Random forest is a machine learning algorithm that can be used for classification and regression tasks. The algorithm operates by building a "forest" of decision trees as shown in Figure 1, where each decision tree is trained on a random subset of features and data from the training set. In simple terms, the random forest algorithm creates multiple decision trees and combines their results to make a prediction. It does this by randomly selecting subsets of features and data from the training set to cre- ate different decision trees. The algorithm then combines the results of these decision trees to make the final pre- diction. Random forest is a popular algorithm because it can handle large datasets with many features and is resistant to overfitting, which is when the algorithm becomes too tuned to the training data and performs poorly on new data. Based on the results received from the algorithm after providing various data regarding the properties of each material, Table 2 shows the ranking of the materials with respect to the criteria shown in this table. Multi-criteria decision analysis The multi-criteria decision analysis (MCDA) is a method used to evaluate and compare multiple options based on a set of criteria. In this case, the criteria used were environmental impact, social responsibility and economic factors. The weight assigned to each criterion was determined based on the importance placed on each factor by decision-makers, i.e., 40 % for environmental impact, 30 % for social responsibility and 30 % for eco- nomic factors. The weight assigned to each criterion plays a crucial role in the material selection process. For example, if the weight assigned to environmental impact was higher than the weight assigned to economic factors, the mate- rial with the lowest environmental impact would be cho- sen even if it was more expensive. Conversely, if the weight assigned to economic factors was higher, the ma- terial with the lowest cost would be chosen regardless of its environmental impact. Based on the ranks and points assigned in Table 3, the MCDA model was used to select the material that best met the criteria set by the decision-makers. The ma- terial with the highest score and rank was selected, tak- P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... 576 Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 Figure 1: Random forest algorithm sample Table 2: Ranking of the materials based on the random forest algorithm Rank Environmental impact Social responsibility Economic factors 1 Mycelium Bamboo Recycled steel 2 Hempcrete Hempcrete Cork 3 Cork Cork Recycled plastics 4 Bamboo Mycelium Natural fibre fabrics 5 Straw bales Natural fibre fabrics Mycelium 6 Recycled steel Recycled steel Bamboo 7 Natural fibre fabrics Straw bales Hempcrete 8 Recycled plastics Recycled plastics Straw bales Table 3: Weightage and points table S. No. Materials Environmental impact points (40 %) Social responsi- bility points (30 %) Economic factors points (30 %) Total points out of 24 Points based on weightage out of 8 1 Mycelium 8 5 4 17 5.9 2 Hempcrete 7 7 2 16 5.5 3 Bamboo 5 8 3 16 5.3 4 Cork 6 6 2 14 4.8 5 Recycled steel 3 3 8 14 4.5 6 Natural fibre fabrics 2 4 5 11 3.5 7 Recycled plastics 11682 . 5 8 Straw bales 42172 . 5 ing into account the weight assigned to each criterion. Using the MCDA, a comprehensive and objective ap- proach was taken to evaluate and compare different ma- terials. 17–19 In short, the MCDA model assigned a weight of 40 % to environmental impact, 30 % to social responsibility and 30 % to economic factors. To select the best material based on the above, points were provided based on the materials rank: Rank 1 has 8 points, Rank 2 has 7 points, Rank 8 has 1 point and other details can be seen in Ta- ble 3. 4 RESULTS As seen in Table 3 we found that mycelium is the best alternative component for air conditioners after ex- amining the eight alternative sustainable materials for air conditioners based on environmental impact, social re- sponsibility and economic factors. Regarding its environ- mental impact, mycelium is a natural, biodegradable and renewable substance; it is a more ecologically friendly alternative than typical insulating materials. Mycelium is produced by fungi from agricultural waste and does not require the use of toxic chemicals. We can lower the car- bon footprint of air conditioners by using this material for air conditioning components. Using mycelium en- courages social responsibility in a variety of ways. First, it promotes sustainable agriculture practises, green jobs and pollution reduction. Second, the usage of mycelium insulating material can help to develop a market for waste agriculture, thus benefiting local farmers and com- munities. Regarding economic factors, mycelium insula- tion manufacture generally operates at low costs, which may greatly cut the air-conditioner production costs. Growing mycelium without the use of artificial fertilisers is both cost-effective and efficient. Mechanical properties of mycelium Mycelium as an alternative material in the sustainable material selection for air conditioners using machine learning and aiming at an environmental impact reduc- tion was chosen based on its excellent mechanical prop- erties, sustainability and cost-effectiveness. Machine learning models, specifically the random forest algo- rithm, were used to optimize the selection of sustainable materials. Mechanical properties of mycelium are shown in Table 4. Table 4: Mechanical properties of mycelium Material Density (kg/m³) Compres- sive strength (MPa) Tensile strength (MPa) Thermal conductivity (W/mK) Mycelium 240–420 0.2–1.4 0.2–35 0.032–0.052 5 DISCUSSION The random forest method is a classification and re- gression analysis supervised machine learning tool. It is a classification method that models a set of decision trees, each of which divides the data into two classes or conditions at each branch. To achieve high accuracy, the algorithm chooses the most essential attributes for each tree and integrates the results from numerous trees. The random forest algorithm may be used in the context of a sustainable material selection for air conditioners to find the most suitable components based on their environ- mental effect, social responsibility and economic vari- ables. The model is trained using historical data that re- late the material and related metrics for mechanical qualities, environmental effect, social responsibility and economic aspects. In comparison to the other options, the algorithm can optimise the use of mycelium in air conditioner components. When selecting mycelium as a suitable material for air conditioners, the random forest algorithm examined the mechanical properties of each material and identified mycelium’s moderate density, good compressive and ten- sile strength and low thermal conductivity, recommend- ing it as a suitable material to be used. Furthermore, when compared to other potential materials, the random forest algorithm considered the cost, sustainability and environmental impact to select the most sustainable and cost-effective option. The application of the random forest algorithm in the context of sustainable material selection for air condi- tioners yields profound implications, particularly in the identification of mycelium as the top choice. The method considered data and attributes related to mechanical qualities, environmental impact, social responsibility and economic factors, contributing to a comprehensive evalu- ation of the materials. Let us delve into the implications of mycelium’s emergence as the optimal material choice: Mechanical Properties: The random forest algo- rithm’s consideration of mechanical attributes is crucial. In the case of mycelium, its moderate density, commend- able compressive and tensile strength and low thermal conductivity render it a feasible option. This highlights mycelium’s potential to fulfil the functional requirements of air conditioner components. Environmental Impact: The algorithm’s capacity to prioritize materials based on their environmental impacts aligns with sustainability objectives. Mycelium’s biodegradability and minimal carbon footprint contribute to its prominence. Its natural growth process and low en- ergy requirements further underscore its eco-friendliness, reducing the overall environmental burden associated with the air conditioner production. Cost-Effectiveness: The integration of economic con- siderations by the algorithm ensures that material choices are not only sustainable but also financially via- ble. Mycelium’s growth process, which can occur in con- trolled environments, may lead to cost savings in the P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 577 long run, as well as potentially reducing waste disposal expenses. Sustainability Criteria: Mycelium’s emergence as the top choice signifies its alignment with a multifaceted set of sustainability criteria. This encompasses not only en- vironmental aspects but also social responsibility and economic viability, demonstrating its ability to contrib- ute to a well-rounded sustainable supply chain. Innovation and Differentiation: The adoption of my- celium as a material for air conditioner components showcases a commitment to innovation and differentia- tion. Such a choice sends a strong signal to consumers and stakeholders about the company’s dedication to re- sponsible practices and its readiness to embrace novel, environmentally conscious alternatives. Industry Influence: The choice of mycelium holds the potential to inspire other manufacturers in the air condi- tioning industry and beyond. As one of the first movers in adopting this innovative material, a company can con- tribute to shifting industry norms toward greener prac- tices. While mycelium holds promise as a sustainable ma- terial for air conditioner components, there are a few bar- riers and challenges that need to be addressed before its widespread adoption in actual manufacturing processes. These challenges include the following: Technical Challenges: Mycelium’s growth process can lead to variations in material properties. Ensuring consistent and reliable material characteristics, such as strength and thermal conductivity, across batches is es- sential for maintaining product quality and performance. Mycelium-based materials might need to undergo rigor- ous testing to ensure they meet durability and longevity expectations, especially in the demanding operational conditions of air conditioners. Supply Chain and Logistics: Scaling up the produc- tion of mycelium-based materials to meet industry de- mand may pose challenges in terms of consistent supply and availability. Logistics related to transporting live my- celium cultures or processed materials might need to be addressed, especially for global supply chains. Waste Management and Disposal: While mycelium is biodegradable, its proper disposal and integration into waste management systems need to be planned to ensure it aligns with the existing disposal infrastructure. In summary, the implications of mycelium’s ranking as the top choice, revealed by the random forest algo- rithm, extend beyond a mere material selection. They signify a step toward enhanced sustainability, responsible resource utilization and innovation within the air condi- tioning sector. By showcasing the multifaceted advan- tages of mycelium, this research not only fosters im- proved decision-making but also bolsters the overall sustainability landscape. 6 CONCLUSIONS In conclusion, optimizing sustainable material selec- tion for air conditioners in the supply chain using ma- chine learning for the environmental impact reduction can be an effective approach for reducing the carbon footprint of air conditioning components. Using machine learning algorithms to evaluate the environmental impact of different materials, manufacturers can identify the most sustainable options for their products. The supply chain’s selection of environmentally friendly materials for air conditioning components may be improved with the use of machine learning algorithms. Machine learn- ing algorithms such as the random forest algorithm can determine the most environmentally friendly solutions while preserving product quality and cutting costs for the producers by analysing enormous volumes of data on the environmental effects of various materials. According to the study, mycelium is the best product for lessening the environmental effect, increasing social responsibility and economic gains for nearby communities, and supporting sustainable agricultural practises. Overall, the applica- tion of machine learning algorithms may help create a more environmentally friendly industry with lower car- bon emissions, more social responsibility and financial gains. In essence, the adoption of mycelium in the manu- facturing of air conditioners presents both opportunities and challenges. Addressing the barriers through ongoing research, technological advancements and collaboration between material scientists, engineers and regulatory bodies will be crucial to realizing the potential benefits of mycelium while navigating the complexities of real-world manufacturing contexts. Future scope Future research on the use of mycelium in air condi- tioners may concentrate on overcoming its drawbacks and enhancing its mechanical qualities while preserving its benefits for ecological and social sustainability. To change the structure and qualities of mycelium to fulfil the specifications for air conditioner components, new approaches need to be developed. Moreover, additional research is required to look into how mycelium insula- tion affects the effectiveness and performance of air con- ditioners. The effects of this material on indoor air qual- ity, noise and energy consumption may be studied and compared to more conventional materials. To ensure the environmental sustainability of mycelium and the sustainability of the mycelium industry, it is also possi- ble to investigate the effects of mycelium production on the ecosystem as a whole. The transition from research to a real-world implementation involves a series of steps, starting with data collection and algorithm training on sustainable material attributes. A validation of the pre- dictions through material testing and collaboration with suppliers helps integrate mycelium-based components seamlessly. Comprehensive lifecycle assessments quan- tify environmental gains. Manufacturers stand to benefit P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... 578 Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 from enhanced sustainability, product differentiation and long-term cost savings. Supply chain managers gain in- sights needed for informed decision-making and opti- mized logistics. Collaboration, pilot projects, continuous improvement and effective market communication are key to transitioning this research into actual sustainable manufacturing practices. However, because mycelium is made of natural fibres, it is sensitive to heat and mois- ture. Consequently, we can use any of the following techniques: coating mycelium with protective films, ther- mal treatment, adding chemicals, combining mycelium with other materials, and optimising growing conditions. 7 REFERENCES 1 A. Gunasegaram, Supply chain management: Theory and applica- tions, European Journal of Operational Research, 159 (2004)2 , 265–268, doi:10.1016/j.ejor.2003.08.015 2 R. Herbrich, M. Keilbach, T. Graepel, P. Bollmann-Sdorra, K. Ober- mayer, Neural networks in economics: Background, applications and new developments, Advances in Computational Economics: Compu- tational Techniques for Modeling Learning in Economics, 11 (2000), 169–196 3 S. M. Ferrandez et al., Optimization of a Truck-Drone in Tandem Delivery Network Using K-Means and Genetic Algorithm, Journal of Industrial Engineering and Management, 9 (2016) 2, 374–388 4 D. Brissaud, S. Tichkiewitch, P. Zwolinski, Innovation in Life Cycle Engineering and Sustainable Development, Conference proceedings, Springer, 2006 5 W. H. Tsai, S. J. Hung, Treatment and recycling system optimisation with activity-based costing in WEEE reverse logistics management: an environmental supply chain perspective, Int. J. Prod. Res., 47 (2009) 19, 5391–5420 6 S. H. Amin, G. Zhang, A three-stage model for closed-loop supply chain configuration under uncertainty, Int. J. Prod. Res., 51 (2013)5, 1405–1425 7 M. Chouinard, S. D’Amours, D. Ait-Kadi, A stochastic program- ming approach for designing supply loops, Int. J. Prod. Econ., 113 (2008) 2, 657–677 8 E. Simangunsong, L. C. Hendry, M. Stevenson, Supply-chain uncer- tainty: a review and theoretical foundation for future research, Int. J. Prod. Res., 50 (2012) 16, 4493–4523 9 A. Jindal, K. S. Sangwan, Closed loop supply chain network design and optimisation using fuzzy mixed integer linear programming model, Int. J. Prod. Res., 52 (2014) 14, 4156–4173 10 E. Ozceylan, T. Paksoy, Interactive fuzzy programming approaches to the strategic and tactical planning of a closed-loop supply chain under uncertainty, Int. J. Prod. Res., 52 (2014) 8, 2363–2387 11 D. Nakandala, H. Lau, J. J. Zhang, Optimization model for transpor- tation planning with demand uncertainties, Ind. Manage. Data Syst., 114 (2014) 8, 1229–1245 12 Y. Zhou et al., Closed-loop supply chain network under oligopolistic competition with multiproducts, uncertain demands, and returns, Math. Prob. Eng., 2014 (2014), 1–15 13 S. Behdad, A. S. Williams, D. Thurston, End-of-life decision making with uncertain product return quantity, J. Mech. Des., 134 (2012)10 14 J. J. Nie et al., Collective recycling responsibility in closed-loop fashion supply chains with a third party: financial sharing or physical sharing?, Math. Probl. Eng., 2013 (2013), 1–11 15 K. K. Pochampally, S. M. Gupta, A multiphase fuzzy logic approach to strategic planning of a reverse supply chain network, IEEE Trans. Electron. Packag. Manuf., 31 (2008) 1, 72–82 16 G. M. Devos Ganga, L. C. Ribeiro Carpinetti, A fuzzy logic ap- proach to supply chain performance management, Int. J. Prod. Econ., 134 (2011) 1, 177–187, doi:10.1016/j.ijpe.2011.06.011 17 G. R. Bivina, M. Parida, Prioritizing pedestrian needs using a multi-criteria decision approach for a sustainable built environment in the Indian context, Environment, Development and Sustainability, 22 (2020), 4929–4950, doi:10.1007/s10668-019-00381-w 18 E. Asadi, Z. Shen, H. Zhou, A. Salman, Y. Li, Risk-informed multi-criteria decision framework for resilience, sustainability and energy analysis of reinforced concrete buildings, Journal of Building Performance Simulation, 13 (2020) 6, 804–823, doi:10.1080/ 19401493.2020.1824016 19 B. Mecca, Assessing the sustainable development: A review of multi-criteria decision analysis for urban and architectural sustain- ability, Journal of Multi Criteria Decision Analysis, (2023), doi:10.1002/mcda.1818 P. SIVARAMAN, S. SANTHOSH: OPTIMIZING SUSTAINABLE MATERIAL SELECTION FOR AIR CONDITIONERS ... Materiali in tehnologije / Materials and technology 57 (2023) 6, 571–579 579