The world’s leading publication for data science, AI, and ML professionals.

Application of an Integrated Life Cycle Assessment and Supply Chain Analytics on LDPE

To analyze sales value and pollution emission data of Single Use Plastics from the United Kingdom using 3 Regression techniques

Notes from Industry

_Authors: Debanjana Chakraborty @chakrabortydebanjana, Suparno Bhatta @iamsuparno and Dr Poovammal E, Department of Computer Science and Engineering, SRM Institute of Science and Technology, Kattankulathur, India, Chennai 603203

Sustainable product refers to the clean green product available at an optimum cost. In the market-dominating era, rapid commercial developments focus on the fast production of inexpensive goods. This has created a surge in over-consumption of natural resources at an expensive rate, thus producing harmful pollutants. On realizing the consequences, nations have started taking measures to eliminate the negative impacts the market is having over the climate and natural environment.

Developing a model that could focus on establishing a balance between environmental quality and the supply chain of a product in the global market has become a matter of utmost importance (Bhatta et al., 2021). This is possible by developing a model which is an amalgamation of the Supply Chain Management Analytics and Life Cycle Assessment (LCA).

Life Cycle Assessment (IS014040 and ISO4044) refers to the method of testing and assessing the environmental and ecological qualities of a product. The entire journey includes the raw material extraction phase, manufacturing and processing phase, transportation phase and finally disposal and reuse phase.

Supply Chain Management (SCM) demands the organization of data, people, events, organizations, and resources (Bhatta et al., 2021). They emphasize improving the functioning, planning, organization and control of the flow of goods. The entire process includes warehousing, supplier, distributor, market, manufacturer and retailer logistics.

DRAWBACKS IN THE EXISTING MODELS:

In the individual models of LCA and SCM, a significant lack in information technology, data generation and management was noticed. (Davenport et al., 2001; Hazen et al., 2014). They were failing to produce an accurate result went tested physically. This is because of a lack of understanding of the key decisional areas necessary to understand product types, quality, and sustainability when used in any specific industry (Fornasiero et al., 2017). The independent models lack logical and technical processes to identify profit margin, consumption and emission patterns (Brondi & Carpanzano, 2011). The data sets used in the previous models are mostly aggregated irrespective of them being location-specific. (Nuasaen et al., 2014; Park & Seo, 2006;Moriguchi & Terazono, 2000).

OBJECTIVE:

A lot of research has taken place to develop individual models of LCA and SCM but very little focus are given on combining both aspects to provide a quality order fulfilment from origin to consumption. The combination of the technologies will establish an operational efficiency thus increasing the productivity of the model. The integrated model will help companies keep a track of their supply-demand by focusing on growth and reduction of production expenditure. Simultaneously, it will also emphasize reducing harmful environmental emissions and over-consumption of natural resources. We have previously proposed an architecture (Bhatta et al., 2021) that will be implemented in this study to check for the efficiency of the model.

Single-Use Plastics
Single-Use Plastics

The entire test will be performed on the product (single-use plastics) that is responsible for major consumption of natural gas and petroleum and emission of harmful greenhouse gasses. The model should emphasize completeness, consistency, accuracy and timeliness, relevancy, value, quantity, accessibility and reputation of data dimensions (Wang, 1996; Wang et al., 1995; Wang, 1998).


IMPLEMENTATION

Architecture Diagram of Integrated Model on LCA and SCM
Architecture Diagram of Integrated Model on LCA and SCM

The model is an integrated development that first identifies the raw materials essential for fulfilling customers’ demand. It then performs LCA and SCM analysis using Python in Google Colab. In the LCA phase, the environmental quality of each raw material is tested to obtain the amount of energy consumed during the process, manufacture and transportation phase. This gives an understanding of the scarcity of the number of non-renewable resources available in the manufacturing phase and the amount of energy consumption that takes place while using these resources. It also gives an understanding of the amount of greenhouse and non-greenhouse emission taking place due to the manufacturing of single-use plastic. Simultaneously the data is also analysed for the supply-demand of the product. This can be obtained by entering the planning phase where the net and gross expenditure in manufacturing and selling a single-use plastic is obtained. The profit and loss are forecasted based on net expenditure, gross expenditure, operational costs and additional third-party costs. Once the data set is ready, the data mining is performed using Scikit learn.

UML Diagram
UML Diagram

· Firstly, we imported the dataset and mounted it on the same drive as that of Google Colab.

mounting dataset to google collab
mounting dataset to google collab

· Then the path was set in python for accessing the dataset.

Adding the pathway of the dataset
Adding the pathway of the dataset

· Then we imported the NumPy Library for converting the data from CSV and to analyse data in table format.

import the NumPy library to convert the dataset into an array of data
import the NumPy library to convert the dataset into an array of data

·All the inputs required were taken using feature extraction and set the target values for prediction.

Setting the data values for input and the output
Setting the data values for input and the output

· The training and testing data were split into 70:30 to obtain the best result.

Train and test the data set
Train and test the data set

· Then we performed Huber, Ridge and Linear Regression after importing Scikit learn.

Code snippet for importing Huber (left),Ridge(Middle) and Linear (Right) Regression
Code snippet for importing Huber (left),Ridge(Middle) and Linear (Right) Regression

· Model scores were checked to obtain the accuracy of the mining technique.

Model Score for Huber (left), Ridge (Middle) and Linear Regression (Right)
Model Score for Huber (left), Ridge (Middle) and Linear Regression (Right)

· For Graph Visualization, the real and predicted values obtained from the target result were compared with Smooth Line Graph. Both LCA and SCM phase is performed using the regression techniques.

Dataset Collection

It is essential to understand the feasibility of the model. Minute datasets have ready-made data sets that are available for testing on an Integrated LCA-SCM model so a new dataset was created for this testing process.

The collection methodology was different for each LCA and SCM sections. The production of single-use plastic is done by low-density polyethylene. This plastics’ data is being used because they presently have the most demand in the global market. Low-Density Polyethylene is very versatile in terms of usage in every aspect of our life. They are also the largest emitters of greenhouse gasses which are responsible for global warming. We obtained the dataset from the first reporting period for the six months (October 2015-April 2016) of single-use carriers submitted to the government of England. This Data (Lca-SCM Database for Plastic-Full Database, n.d.) includes net Expenditure in manufacturing the plastic, gross expenditure in selling the plastic, VAT by Govt.

We combined the SCM data with a self-created dataset for LCA. Values were calculated using ‘No. of Single-use Plastic’ and the value for one unit production or emission based on values given by a technical report on Cradle-to-Gate Life-Cycle Inventory of Nine Plastic Resins & Four Polyurethane Precursors August 2011(Feraldi et al., 2013). The dataset for LCA focuses on the ‘Cradle to Gate’ phase. This data includes values of Non-renewable resources used and Energy consumed during the process. It also includes various Greenhouse and Non-greenhouse pollutants emitted into the atmosphere during the manufacturing, process and transportation phase of single-use plastics.

To get the dataset please mail the author of this publication

Regression Techniques

We used Linear Regression as it helps in predicting the net profit/loss of a company depending on the sale of the raw materials for manufacturing purpose. It gives accurate results for predictive analysis where we use one dependent and one independent experimental variable. As it’s good for forecasting effect, and determining the strength of the predictor. We also tried using a couple of more regression techniques to check which one is better suited for us. Huber Regression was used as it is robust to outliers. It uses a special loss function compared to the standard least square method. It seems usual to possess the smallest amount squares penalty for small residuals of loss but on big residuals, its penalty is lesser and it linearly surges instead of quadratically. Hence, it works well for datasets having enormous outliers.

Then we used Ridge Regression as it analyses multiple correlation data that has severe multi-collinearity. The least-squares method estimates are unbiased when multicollinearity occurs but their variances are huge which will mean that it’s going to be far away from truth value. It reduces the quality errors by adding a degree of bias to the regression estimates.


RESULTS AND DISCUSSIONS

We wanted to design a model that could predict a trend between the environmental quality and sustainable cost of a product. In this result, we tried to find the rate of energy consumption, sales factor, and harmful pollutants that are emitted during the manufacturing of single-use plastic. During this study Linear, Huber, and Ridge Regression techniques were implemented for a comparative study to know which method would work the best during this case in terms of accuracy.In the comparative study, actual and real values are calculated from the graph.The actual curve represents the present status of the curve based on independent variables.The predicted curve represents the future predictions of how the curve might change in comparison to present values.

In all the three regression techniques, the model score nearly ranged from 0.9995–1.0. This shows the level of accuracy of the three techniques. All three regression techniques provided us with almost similar results with micro deviations. However, out of the three Huber Regression showed the highest accuracy with a model score of 1.0 in almost all the codes. Over here we will only show graphs and data heads of Huber regression.

SALES AND INFERENCE:

We are trying to find the amount of gross expenditure done by companies in the UK annually to manufacture single-use plastics from low-density polyethylene. In the dataset, we have been provided with values for net expenditure, gross expenditure and VAT.

Based upon Calculation: Gross Proceed = Net proceed + VAT (Gross>>Net Proceed)

Gross Proceed: Average Expenditure of the company in Manufacturing single-use plastic.

Net Proceed: Average Expenditure of the manufacturer in making the Plastic.

VAT: Government Tax on Single-Use Plastic

The following figures illustrate:

Growth of sales of Single-use plastic with and without VAT
Growth of sales of Single-use plastic with and without VAT
  1. In our graph, we notice that the predicted gross value is slightly lesser than the real gross value. There will be a decrease in future expenditure in manufacturing plastics. This means that UK companies will start decreasing their use and manufacturing of these plastics thus adhering to the Anti Plastic Pact signed by the companies in the UK. Companies are yet to cut down on the usage of single-use plastic. A significant decrease in the future production of single-use plastics will help reduce the sale to a significant extent. Companies are yet to cut down on the usage of single-use plastic.
  2. The actual gross expenditure and the predicted gross expenditure in the above all four graphs remains the same. This shows that the government enforced mechanisms have not yet shown their effect. The annual supply and sales of plastic in future will continue to keep increasing among the companies.
Head of top 7 values in Sales without VAT (left), Head of top 7 values in Sales with VAT (right)
Head of top 7 values in Sales without VAT (left), Head of top 7 values in Sales with VAT (right)
  1. The two data sets in the above figure are compared to understand the minute change in the predicted gross expenditure. Micro changes in predicted gross in the data head are nearly indistinguishable to the naked eye. The data set with VAT shows less predicted expenditure on plastic, while the one without VAT depicts a linear increase in the future sales and consumption of single-use plastic. Thus showing the role of government in controlling the production of single-use of plastic. The strict mechanism enforced by the UK government will force the companies to automatically reduce the manufacturing and consumption of plastic.

ENERGY AND RESOURCE CONSUMPTION:

Chemical Structure of Low-Density Polyethylene
Chemical Structure of Low-Density Polyethylene

Single-use plastic is made up of low-density polyethylene. They are processed from crude oil or natural gas. During the manufacturing of plastics, a lot of non-renewable resources are required for the extraction of raw materials, processing these raw materials to low-density polyethylene and then transporting them from factories to companies. The energy consumption that takes places during the process, transport and material phase helps us understand the amount of non-renewable resources that are being depleted by companies annually.

Process Transportation and Material Phase
Process Transportation and Material Phase

MATERIAL PHASE:

The following figures illustrate: (add figures)

Actual (left) Material Resources vs All non-renewable Energy and Predicted (right) Material Resources vs All non-renewable Energy in Huber Regression
Actual (left) Material Resources vs All non-renewable Energy and Predicted (right) Material Resources vs All non-renewable Energy in Huber Regression
  1. The above figures depict the excessive level of over-consumption of the depleting natural resources. A large amount of natural gas is processed and consumed to prepare low-density polyethylene. Petroleum comes second in position with total energy consumption. Very little energy is consumed from the remaining resources (coal, nuclear, others).
  2. In both the graphs the energy consumption for all resources continues to steeply increase. This depicts the harmful habits of companies to over-consume non-renewable resources.

PROCESS PHASE

The following figures illustrate:

Actual Process Resource vs All non-renewable Energy (left) and Predicted Process Resource vs All non-renewable Energy (right) in Huber Regression
Actual Process Resource vs All non-renewable Energy (left) and Predicted Process Resource vs All non-renewable Energy (right) in Huber Regression
  1. In the above figures, we can see that natural gas consumes maximum energy to prepare low-density polyethylene. Similar to the material phase, petroleum comes second with total energy consumption. The curve for other gasses has almost merged in the graph as they consume very minimal energy which is almost invisible to the eye.
  2. The consumption curve for the process phase is similar to the material phase. This tells us how companies continue to over-consume resources

TRANSPORTATION PHASE:

The following figures illustrate:

Actual Transportation Resource vs All non-renewable Energy (left) and Predicted Transportation Resource vs All non-renewable Energy (right) in Huber Regression
Actual Transportation Resource vs All non-renewable Energy (left) and Predicted Transportation Resource vs All non-renewable Energy (right) in Huber Regression
  1. Once again natural gas and petroleum bag the first two position for the energy consumption in the transportation phase. However, there is a huge difference in the curve for the energy consumption of petroleum and natural gas. This indicates that in comparison to petroleum, natural gas is consumed in much more amount. This was not the case for the process and manufacturing phase.

COMBINE RESOURCES:

The figure of the data head given below represents a comparison between energy consumed by non-renewable resources vs real and predicted values of process, transport and material resources.

The following figures illustrate:

Head of top 9 values from combined Process, Transportation and Material Phase
Head of top 9 values from combined Process, Transportation and Material Phase

6.Energy consumption from non-renewable resources is maximum during the manufacturing of the plastic(process phase). Transportation of raw materials (transportation phase)comes second and the extraction of raw materials to manufacture plastics comes in the third position

7.The companies in the UK show an increase in the future trend of the production of more plastic. This is evident as the predicted process values are more than the real process values. The only way to reduce production is by increasing the tax rate per production. This will create a demand reduction thus reducing excessive use of petroleum and natural gas.

  1. An opposite trend is noticed in the real transportation value and predicted value of transportation. This is opposite to the process phase. This shows that the rate of energy consumption in the transportation phase will decrease in future.

9.The energy consumption in the material phase for both predicted and real value is nearly negligible when compared to values from the production and transportation phase.

The greater the energy consumption, the greater is the depletion of natural resources. Thus, affecting the ecological factor of the planet. It is also increasing the rate of carbon dioxide and methane emission which is responsible for global warming.

GREENHOUSE AND NON-GREENHOUSE EMISSION INFERENCE:

Previously we have noticed how non-renewable resources are being over consumed to manufacture single using plastics. The Pie Chart in figure belowand 15 analyze the greenhouse and non-greenhouse gas emitted during the phase.

GREENHOUSE GASSES:

The figure below analyses the amount of greenhouse gas that is predicted to be emitted when the single-use plastic is extracted, manufactured and transported.

Greenhouse Gasses: Carbon Dioxide, Methane, Nitrous Oxide, Methyl Bromide, Methyl Chloride, Trichloroethane, Chloroform, Carbon tetrachloride, CFC13, HCFC 22.

The following figures illustrate:

Predicted **** Amount of Greenhouse Gases Emitted during all phases (in 1000 Kg)
Predicted **** Amount of Greenhouse Gases Emitted during all phases (in 1000 Kg)
  1. Carbon Dioxide (78.67%) and Methane (20.88%) are produced the most followed by other pollutants. This is because petroleum and natural gas resources are highly over consumed to manufacture single-use plastics. In the transportation phase, a lot of coal along with petroleum and natural gas is also used as fuel. They are the major sources of trapped methane gas. When these resources are over consumed, deadly methane and Carbon dioxide are released into the air. In the transportation phase, a lot of coal along with petroleum and natural gas is also used as fuel. This releases harmful pollutants. Both the pollutants capture heat and don’t allow them to escape from the earth’s surface. Thus making the air extremely hot and increasing the danger of climatic tipping point. This will have a devastating effect on the climate of the planet.
  2. Methane is a deadlier greenhouse pollutant as it remains in the atmosphere for a longer time than Carbon Dioxide. The steep rise in the curve in all three phases can cause many health issues like vision problems, memory loss, nausea, vomiting, facial flushing and headache.
  3. The remaining pollutants together constitute only a micro amount of 0.44 % of the total pollution. So on controlling the rate for methane and Carbon Dioxide, then the planet’s temperature will be controlled.

NON-GREENHOUSE GASSES:

A similar Pie Chart figure has been made below to compare the levels of non-greenhouse gasses that are emitted during the process, transportation and material phase. These gases are not responsible for global warming but they are responsible for contamination of air causing deadly health issues.

Non-Greenhouse Gasses: Particulate Matter (PM unspecified, 2.5, 10), Volatile Organic Compound (VOC), Carbon Dioxide non-fossil, nitrogen oxide, carbon monoxide and Sulphur dioxide. All three phases show the same trend.

The following figures illustrate:

Predicted **** Amount of Non-Greenhouse Gases Emitted during all phases (in 1000 Kg)
Predicted **** Amount of Non-Greenhouse Gases Emitted during all phases (in 1000 Kg)
  1. Approximately 43.5% of Sulphur dioxide is produced during the process. This is an alarming issue as such high percentage damages the ecological system and also responsible for deteriorating heritage buildings. They also cause shortness of breath and chest tightness in humans.
  2. Nitrogen Oxides forms 27% of the total non-greenhouse gasses that are emitted. They are responsible for damaging the vegetation, affecting animals and human beings. This pollutant cause Eye irritation, lung problem and breathing issue. They combine with Sulphur Dioxide which is also emitted in abundance to produce acid rain. Acid rain chemically reacts with buildings and statues and dissolve the chemicals thus making them weak and brittle.
  3. In comparison to others, Carbon monoxide and VOC are produced at a decent rate of 12% Carbon monoxide is still as carbon monoxide above 40% these pollutants become lethal to human and animal lives.
  4. Particulate Matter (unspecified, 10 and 2.5) are emitted in very micro amounts in comparison to the other pollutants These are majorly micro and nano fine particles that are the main contributors to indoor pollution. They have a significant impact on human health. These are a complex mixture of extremely small particles and liquid droplets. These pollutants are deadly in nature and are responsible for even deaths when emitted at a huge level.

Carbon Dioxide fossil emission is almost negligible. However, they are less deadly than the other pollutants if produced in high concentration can lead to coma or death.


CONCLUSION

All three regression techniques were tested for accuracy. The model that gives the best accuracy is Huber followed by Linear and then Ridge. This is because it is less sensitive towards outliers or abnormalities in data that is present in our data set. However, during the process, certain net, gross, and vat values were not available, so they were assumed 0 by default. Availability of complete data has been an issue for ages and it was also reflected during this study. Most of these data are hidden or unexplored due to confidentiality reasons of companies. In most of the graphs, all the curves are not clearly visible. This is because our dataset contains a large range of values which when cumulated together becomes difficult for us to see the minute difference in result through our naked eye. The data used in this process is independent of any specific location so the results might vary if any specific region is taken into consideration. The entire process using the regression technique has provided very high-efficiency results. The readers could try clustering or classification techniques as well to obtain a sustainable result.


REFERENCES:

Bhatta, S., Chakraborty, D., & Poovammal, E. (2021). ARRIVING AT OPTIMAL COST FOR A SUSTAINABLE PRODUCT USING AN INTEGRATED LCA-SCM MODEL. 9(2), 1193–1199. https://doi.org/10.17762/itii.v9i2.472

Brondi, C., & Carpanzano, E. (2011). A modular framework for the LCA-based simulation of production systems. CIRP Journal of Manufacturing Science and Technology, 4(3), 305–312. https://doi.org/10.1016/j.cirpj.2011.06.006

Davenport, T. H., Harris, J. G., De Long, D. W., & Jacobson, A. L. (2001). Data to Knowledge to Results. California Management Review, 43(2), 117. http://search.ebscohost.com/login.aspx?direct=true&db=buh&AN=4372997&site=ehost-live

Feraldi, R., Cashman, S., Huff, M., & Raahauge, L. (2013). Comparative LCA of treatment options for US scrap tires:

Material recycling and tire-derived fuel combustion. International Journal of Life Cycle Assessment, 18(3), 613–625. https://doi.org/10.1007/s11367-012-0514-8

Fornasiero, R., Brondi, C., & Collatina, D. (2017). Proposing an integrated LCA-SCM model to evaluate the sustainability of customisation strategies. International Journal of Computer Integrated Manufacturing, 00(00), 1–14. https://doi.org/10.1080/0951192X.2016.1268716

Hazen, B. T., Boone, C. A., Ezell, J. D., & Jones-Farmer, L. A. (2014). Data quality for data science, predictive analytics, and big data in supply chain management: An introduction to the problem and suggestions for research and applications. International Journal of Production Economics, 154, 72–80. https://doi.org/10.1016/j.ijpe.2014.04.018

lca-scm database for plastic-full database. (n.d.).

Moriguchi, Y., & Terazono, A. (2000). A simplified model for spatially differentiated impact assessment of air emissions. International Journal of Life Cycle Assessment, 5(5), 281–286. https://doi.org/10.1007/BF02977580

Nuasaen, S., Opaprakasit, P., & Tangboriboonrat, P. (2014). Hollow latex particles functionalized with chitosan for the removal of formaldehyde from indoor air. Carbohydrate Polymers, 101(1), 179–187. https://doi.org/10.1016/j.carbpol.2013.09.059

Park, J. H., & Seo, K. K. (2006). A knowledge-based approximate life cycle assessment system for evaluating environmental impacts of product design alternatives in a collaborative design environment. Advanced Engineering Informatics, 20(2), 147–154. https://doi.org/10.1016/j.aei.2005.09.003

Wang, R. Y. (1996). Beyond accuracy: What data quality means to data consumers. Journal of Management Information Systems, 12(4), 5–34. https://doi.org/10.1080/07421222.1996.11518099

Wang, R. Y. (1998). A Product Perspective on Total Data Quality Management. Communications of the ACM, 41(2), 58–65. https://doi.org/10.1145/269012.269022

Wang, R. Y., Storey, V. C., & Firth, C. P. (1995). A Framework for Analysis of Data Quality Research. IEEE Transactions on Knowledge and Data Engineering, 7(4), 623–640. https://doi.org/10.1109/69.404034


Related Articles