Revista de Ciencias Tecnológicas (RECIT). Volumen 3 (1): 10-22

Revista de Ciencias Tecnológicas (RECIT). Universidad Autónoma de Baja California ISSN 2594-1925

Volumen 9 (2): e420. Abril-Junio, 2026. hps://doi.org/10.37636/recit.v9n2e420

ISSN: 2594-1925

Research article

Effect of a Biodegradable Additive on the Mechanical

Properties of LDPE Blown Films: A Statistical and

Machine Learning Approach

Efecto de un aditivo biodegradable en las propiedades mecánicas de películas sopladas

de LDPE: un enfoque estadístico y de aprendizaje automático

Gilberto Alarcón Aguilar1, Alam Josué Reyes López2, Frixia Galán-Méndez1*

1Universidad Veracruzana, Facultad de Ciencias Químicas, Circuito Gonzalo Aguirre Beltrán s/n, zona

Universitaria, Xalapa, Veracruz, México, C. P. 91000.

2Universidad Veracruzana, Maestría en Ingeniería de la Calidad, Circuito Gonzalo Aguirre Beltrán s/n,

zona Universitaria, Xalapa, Veracruz, México, C. P. 91000.

Autor de correspondencia: Frixia Galán-Méndez, Universidad Veracruzana, Facultad de Ciencias Químicas,

Circuito Gonzalo Aguirre Beltrán s/n, zona Universitaria, Xalapa, Veracruz, México, C. P. 91000. Correo

electrónico: fgalan@uv.mx. ORCID:

Recibido: 5 de Julio del 2025 Aceptado: 19 de Junio del 2026 Publicado: 24 de Junio del 2026

Resumen. - El estudio evalúa el efecto que tiene la adición del compuesto biodegradable P–Life sobre las

propiedades mecánicas de bolsas para hielo “IB”, fabricadas a partir de polietileno de baja densidad (LDPE). Se

recolectaron datos de resistencia a la tracción, elongación, resistencia al rasgado, resistencia al punzonado e

impacto durante un periodo de seis meses en una planta manufacturera. Se estudiaron correlaciones entre las

propiedades mecánicas y las formulaciones con (B) y sin aditivo biodegradable (NB). El análisis de Pearson mostró

correlaciones positivas entre el grosor de la película y sus propiedades (r ≥ 0.5). El ANOVA reveló diferencias

significativas (p ≤ 0.05) entre ambas formulaciones en algunas propiedades. Además, se desarrolló un modelo

predictivo mediante el algoritmo M5P con una precisión del 94.147%, validado con muestras reales. Los resultados

sugieren que el uso de inteligencia artificial es viable para predecir propiedades mecánicas en procesos de

extrusión de polímeros, lo que puede optimizar el control de calidad industrial.

Palabras clave: Extrusión soplada; Propiedades mecánicas; Polietileno; ANOVA; Aprendizaje automático.

Abstract.- This study evaluates the effect of incorporating the biodegradable additive P–Life on the mechanical

properties of ice bags (IB) made from low-density polyethylene (LDPE). Mechanical resistance to tension,

elongation, tearing, puncture, and impact was measured over a six-month production period. Correlations were

analyzed between mechanical performance and the presence or absence of the additive. Pearson’s coefficient

indicated positive correlations (r ≥ 0.5) between film thickness and mechanical properties. ANOVA revealed

statistically significant differences (p ≤ 0.05) in several properties depending on the formulation. In addition, a

predictive model based on the M5P algorithm was developed, achieving 94.15% accuracy, and was validated with

actual samples. The results suggest that machine learning is a viable tool for predicting mechanical behavior in

polymer extrusion processes, offering improvements in industrial quality control.

Keywords: Blown film extrusion; Mechanical properties; Low-density polyethylene; ANOVA; Machine learning.

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

1. Introduction

In 2020, the food industry generated 826,000 tons of packaging, of which 239,000 tons were allocated to

disposable products [1], [2]. Most of this packaging is manufactured with non-biodegradable plastics such

as LDPE, resulting in considerable environmental impact [3], [4]. To address this problem, additives like

P-Life—derived from coconut palm oil—have been developed to accelerate plastic degradation [5], [6].

Previous studies have demonstrated that these additives can modify the mechanical properties of plastic

films depending on their molecular interactions [7], [8],[9]. Furthermore, the operating conditions of the

extrusion process significantly influence these properties [10].

Despite advancements, modeling polymer transformation processes remains complex due to nonlinear

relationships between variables [11], [12]. The use of machine learning tools, such as M5P algorithms,

offers an alternative for predicting properties without resorting to costly experimental methods [13]–[16].

For instance, in the concrete industry, mechanical properties were predicted using the M5P artificial

intelligence algorithm with approximately 97% accuracy [17].

Similarly, it has been employed in numerous materials science scenarios, facilitating the prediction of

physical properties under temperature variations [18]. Additionally, this technology is also utilized to

optimize processes where numerical data availability is not limited [19].

The M5P artificial intelligence algorithm excels particularly in predicting mechanical properties, unlike

conventional regression methods [20]. Therefore, the objective of this work is to demonstrate the potential

of the M5P algorithm for predicting the mechanical properties of LDPE films and to statistically evaluate

the difference between formulations with and without biodegradable additives, providing evidence to

support industrial implementation.

2. Materials and methods

2.1 Materials

Low-density polyethylene (LDPE) supplied by Braskem Indesa was used, characterized by its high

mechanical strength and versatility for processes such as blown extrusion and injection molding (density:

0.932 g/cm³; melt flow index: 0.25 g/10 min at 190 °C/2.16 kg).

The biodegradable additive employed was P–Life, commercial name SMC 100, dosed at 1–3% by mass.

This additive is designed for plastic films and exhibits a density of 1.2 g/cm³ and a melt flow index of 2–

10 g/10 min.

2.2 Equipment used

An industrial TECOM extruder (OLGIATE OLONA, Italy) was employed, configured with a linear

temperature profile of 185 °C in all zones, a mass flow rate of 45 kg/h, and a screw speed of 120 rpm.

For mechanical tests, a GBD-2 universal testing machine (Electronic Tensile Tester, China) was used,

equipped with pneumatic grips at 6 bar pressure (Figure 1).

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

For impact testing, a GBD-12 universal testing machine (Falling Dart Impact Tester, China) was utilized,

equipped with pneumatic holders at 6 bar pressure (Figure 2).

Figure 1. Universal testing machine GBD-2 Electronic

Tensile Tester. (Source: Own elaboration).

Figure 2. Impact machine GBD-12 Falling Dart Impact

Tester. (Source: Own elaboration).

2.3 Mechanical characterization

Samples were extracted from rolls intended for ice bag (IB) production, evaluated exclusively in the

machine direction (MD) as established by the Mexican standard NMX-134-CNCP-2013. (Figure 3). Tests

were conducted at room temperature (23 ± 2 °C), in accordance with the corresponding Mexican

standards.

Figure 3. Extrusion direction of polyethylene film. (Source: own elaboration).

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

2.3.1 Tensile strength and elongation

Five test pieces were cut from the extruded film, their dimensions are 16 mm wide and 100 mm long,

subsequently the testing machine was set to 100 mm/min with a separation between jaws of 50 mm (Figure

4), in accordance with the Mexican standard NMX-134-CNCP-2013.

2.3.2 Tearing and Puncture Resistance

Five specimens were cut per test. Tests were performed at 50 mm/min. For the tear, the grip separation

was 25 mm (Figure 4), in accordance with Mexican standard NMX-E-112-CNCP-2014. In the puncture

test, the needle made surface contact with the specimen, measuring 100 mm wide and 100 mm long.

Figure 4. Specimen used in tearing resistance testing

2.3.3 Impact resistance

Eight test specimens were cut per test, each measuring 100 mm wide and 100 mm long. The dart was

dropped from a height of 660 mm in free fall, impacting the specimen (Figure 7), in accordance with

Mexican standard NMX-E-099-CNCP-2014.

2.4 Statistical analysis

Data was collected over six months, including operational conditions (extrusion, ambient, and coil

temperatures), thickness, and mechanical properties (tensile strength, elongation, tearing, puncture, and

impact resistance). The analysis consisted of two phases: 1) Exploration of correlations using Pearson’s

coefficient. 2) Analysis of variance (ANOVA) to evaluate the effect of formulation type, with thickness as

a blocking factor. The assumptions of normality, homoscedasticity, and independence were validated

through graphical residual analysis.

3 Results and discussion

Operating variables—particularly temperature and screw speed—significantly influence (p ≤ 0.05) the

mechanical properties of films produced by extrusion [7], [8]. In this context, Table 1 presents the key

variables monitored during the blown extrusion process for manufacturing ice bags made from low-

density polyethylene (LDPE).

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

Table 1. Variables in the blow molding extrusion process.

Variable

Lower limit

Upper limit

Range

Operating temperature (°C)

180

195

Operating pressure (bar)

350

N.A.

Upper roller pressure (bar)

Screw speed (rpm)

120

N.A.

Cooling (%)

Furthermore, Figure 8 shows the flow diagram of the blown extrusion process for polyethylene film

formation. This diagram enables identification of the critical variables previously mentioned, which were

considered for conducting the statistical analysis described in the following sections.

Table 2. Correlation of variables with the through the Pearson method.

Variables

Caliber

(ga)

Calibration

(%)

coil

(°C)

extruder

(°C)

Ambient

temperature

(°C)

Tensile

strength

(Mpa)

Elongation

(%)

torn

(kgF)

punching

(N)

Calibration

(%)

-0.042

T. Coil (°C)

-0.134

0.067

T. extruder

(°C)

-0.081

0.277

0.302

Ambient

temperature

(°C)

-0.150

-0.028

0.768

-0.119

Tensile

strength

(Mpa)

0.869

-0.066

0.131

0.006

-0.237

Elongation

(%)

0.685

-0.152

0.078

-0.181

-0.111

0.692

R. torn

(kgF)

0.860

-0.015

0.213

0.109

-0.180

0.778

0.498

R. punching

(N)

0.786

-0.019

0.375

0.139

-0.375

0.764

0.489

0.708

Impact (g)

0.533

0.083

0.085

0.034

-0.257

0.527

0.543

0.407

0.533

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

Figure 5. Flow diagram of the blown extrusion.

Based on 40 observations—which included the

independent variables of actual thickness,

thickness deviation, and average temperatures of

the coil, extruder, and ambient, along with the

mechanical properties of tensile strength,

elongation, tearing resistance, puncture

resistance, and impact resistance—the Pearson

correlation coefficient was calculated using

Minitab 2021, as shown in Table 2. We selected

Pearson's correlation coefficient to demonstrate

potential relationships between variables, as

reported by Temizhan et al. [21], since this

methodology is highly efficient for revealing

correlations in real-world data.

The results show (r ≥ 0.5) a positive and

significant correlation between actual thickness,

and all evaluated mechanical properties (tensile

strength, elongation, tearing resistance, puncture

resistance, and impact resistance). Likewise, a

significant association was detected between

ambient temperature and coil temperature—a

phenomenon consistent with the second law of

thermodynamics: heat flows from the higher-

temperature system (coil) toward the lower-

temperature system (ambient) until equilibrium

is reached, driven by the existing thermal

gradient. Tensile strength, in turn, exhibited

positive correlations with the remaining

mechanical properties. The direct relationship

with elongation is particularly notable: the higher

the maximum force sustained before rupture, the

greater the unit deformation recorded by the film

reflecting a more ductile polymer matrix.

Analogous reasoning extends to tearing,

puncture, and impact tests, which share the

principle of subjecting the film to separation

stresses until failure. Finally, elongation showed

a positive correlation with impact resistance.

During free-fall impact testing, the film

undergoes extensive deformation; consequently,

high elongation performance translates

predictably into greater energy absorption

capacity before fracture.

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

These findings confirm that the analyzed mechanical properties are strongly interrelated and underscore

the importance of controlling thickness and thermal process variables to ensure consistent mechanical

performance in extruded LDPE films.

3.1 Analysis of variance with thickness as a blocking factor

To identify statistically significant differences in mechanical properties based on formulation, an analysis

of variance (ANOVA) was performed using a set of 50 observations. Thickness was incorporated as a

blocking factor to control its influence on system variability. Box-Cox transformations were applied to the

response variables to better satisfy ANOVA's assumptions of normality and homogeneity of variances by

Atkinson, et al. [22].

The analysis began by evaluating the effect of "thickness" and "formulation" factors on tensile strength,

employing a Box-Cox transformation with parameter λ = 2. Subsequently, variance analysis was

conducted using the transformed variable as the response.

Table 3 shows the ANOVA results, demonstrating that thickness has a significant effect on the film's tensile

strength (p ≤ 0.05), while no statistically significant differences attributable to the formulation were

observed.

Table 3. Analysis of variance for response tensile strength transform.

Source

SC Adjustment

MC Adjustment

F value

P-value

Biodegradable

16749

0.19

0.667

Caliber

60620038

8660005

96.08

0.000

Error

241

21721776

90132

Total

249

82355274

The residual analysis indicated that the statistical assumptions required for model validity were met as

residuals are normally distributed, centered around zero, exhibit no heteroscedasticity patterns, and show

no autocorrelation. Consequently, the assumptions of normality, homoscedasticity, and independence were

verified. Additionally, a post hoc multiple comparison test using Fisher's method (LSD) was applied to

explore in greater detail the differences between the levels of the identified significant factor, as

demonstrated by Lu et al., [23], these results are shown in Table 4.

Table 4. Fisher's LDS method and a 95% confidence (tensile strength).

Caliber (ga)

Average (MPa)

Group

300

54.009

220

37.770

190

37.445

200

35.242

210

32.613

175

30.449

180

29.644

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

170

26.022

The results show an upward trend in tensile strength as thickness increases. The lowest strength was

recorded in thin films, with highly significant differences between the 170 ga thickness (26.0 MPa) and

the 300 ga thickness (54.0 MPa). This behavior aligns with the Pearson correlation analysis, which showed

a direct association between thickness and mechanical properties.

When comparing these values with those reported by Mallegni et al. [24]—who analyzed a plasticized

PLA/PBAT film of 0.05 mm (≈200 ga) with a strength of 25 MPa—it is observed that a film of identical

thickness manufactured primarily from low-density polyethylene achieves a substantially higher mean

strength (35.2 MPa). This difference can be attributed to both material composition and process

conditions: in the present study the extruder operated at 120 rpm, whereas Mallegni et al. [24] employed

200 rpm; this influence of screw speed on mechanical properties was previously documented by Gálvez

et al. [8].

To analyze the effect of thickness and formulation on elongation, a Box-Cox transformation with λ = 3

was applied before conducting the ANOVA, ensuring compliance with normality and homogeneity of

variances assumptions.

Table 5 presents the ANOVA results for the elongation variable, showing that thickness constitutes a

statistically significant factor (p ≤ 0.05), while no significant effects attributable to the formulation were

detected. Residual analysis confirmed model assumption compliance: errors are normally distributed,

centered around zero, with no relevant violations of homoscedasticity or evidence of autocorrelation,

validating residual independence.

Table 5. Analysis of variance for response elongation transform.

Source

SC Adjustment

MC Adjustment

F value

P-value

Biodegradable

165

0.04

0.841

Caliber

1173067

167581

40.82

0.000

Error

241

959391

4105

Total

249

2168610

Subsequently, Fisher's post hoc test (LSD) was applied to perform multiple comparisons between the

levels of the significant factor, the results of which are presented in Table 6.

Table 6. Fisher's LDS method and a 95% confidence (elongation).

Caliber (ga)

Average (%)

Group

300

832.725

190

737.113

210

691.023

200

688.613

220

678.024

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

180

100

633.386

175

612.068

170

611.669

The data reveals a direct relationship between film thickness and elongation: greater thickness yields

higher elongation resistance. The thinnest films showed the lowest elongation values, with a statistically

significant difference between the 170 ga thickness (611.67%) and 300 ga thickness (832.73%). This

behavior is primarily attributed to increased film thickness, which enhances material stretching during

extrusion—consistent with the Pearson correlation analysis showing a positive association between

thickness and elongation.

These results can be compared to those reported by Aliotta et al. [25], who documented elongation

percentages of 48.77%, 107.94%, and 295.8% for specimens averaging 160 ga thickness. Contrasted with

this study’s average for 170 ga films (the thinnest set) of 611.67%, a notable difference emerges. Although

formulations differ, the results indicate that low-density polyethylene (LDPE) films exhibit higher

ductility than those composed of polylactic acid (PLA) and polybutylene succinate adipate (PBSA).

Another relevant factor is extruder screw speed. The comparative study operated at 300 rpm—

significantly higher than the 120-rpm used here. According to Gálvez et al. [8], increased rpm can

adversely affect mechanical properties, suggesting observed differences may also be influenced by this

operating parameter.

Complementarily, Mallegni et al. [24] reported average elongation near 250% for a 200 ga film composed

primarily of PLA/PBAT with a plasticizer (Ej400). In this study, the same thickness (200 ga) achieved an

average elongation of 688.61%, reinforcing LDPE’s superior deformability. The plasticizer in comparative

study likely impaired mechanical properties, reducing elongation capacity.

Additionally, Gómez-Bachar et al. [26], reported elongation in the range of 170–530% for films composed

primarily of starch with an approximate thickness of 250 ga, demonstrating that biodegradable compounds

can be equally resistant as non-biodegradable ones according to Itabatana et al. [11].

Also, Kim et al. [27], reported that the addition of biodegradable compound based on rice husks in 0.5%

composition is as resistant as the extruded film from LDPE, [11].

Furthermore, the effect of "thickness" and "formulation" factors on tearing resistance was analyzed. A

Box-Cox transformation with parameter λ = 0.5 was applied before conducting the corresponding

ANOVA.

Table 7 presents the ANOVA results for film tearing resistance. The data confirms that thickness is a

statistically significant factor (p ≤ 0.05) for this mechanical property, while no significant differences

attributable to formulation were identified.

Table 7. Analysis of variance for response tear strength transformation.

Source

SC Adjustment

MC Adjustment

F value

P-value

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

Biodegradable

0.00233

0.002325

1.19

0.275

Caliber

0.80578

0.115112

59.14

0.000

Error

241

0.46908

0.001946

Total

249

1.2749

Residual analysis confirmed compliance with the statistical model assumptions: errors are normally

distributed, exhibit no heteroscedasticity, and are independent, thereby validating the ANOVA model

application. Based on these results, a multiple comparisons test was performed using Fisher's method

(LSD), the results of which are presented in Table 8.

Table 8. Fisher's LDS method and a 95% confidence (tear strength).

Caliber (ga)

Average (kgF)

Group

300

0.871318

220

0.729295

210

0.617177

200

0.605968

190

0.541330

170

0.532694

180

100

0.531749

175

0.518209

The results show a direct relationship between film thickness and tearing resistance: greater thickness

requires higher force to induce tearing fracture. The thinnest films exhibited the lowest resistance values,

with a statistically significant difference between the 170 ga thickness (0.532 kgf) and 300 ga thickness

(0.871 kgf).

This behavior is primarily attributed to extruded film thickness, which directly influences structural

integrity. As evidenced by Pearson correlation analysis, a positive linear relationship exists between

thickness and tearing resistance. Consequently, increased material thickness systematically enhances its

capacity to withstand shear stresses—consistent with expected behavior for extrusion-processed polymer

films.

These results can be compared to Aliotta et al. [25] and Mallegni et al. [24], who documented tearing

resistances of 5.24 N/mm for 200 ga films and 4.69 N/mm for 160 ga films, respectively. Although

formulations differ, thick variations likely explain observed differences. For comparison, the force per

millimeter was calculated from reported values, yielding 0.0267 kgf and 0.0189 kgf, respectively. These

contrast with this study's results for 170 ga and 200 ga thicknesses (0.5326 kgf and 0.6059 kgf), confirming

polyethylene films exhibit superior tearing resistance compared to primarily polylactic acid (PLA)-based

films [28], [29].

This significant difference can be attributed to multiple factors: extrusion speed, formulation, and

experimental conditions. Crucially, this study's tests followed the Mexican standard NMX-E-112-CNCP-

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

2014—applicable in the company's jurisdiction—hence results may differ from those obtained under

region-specific regulations.

This contrast is supported by Zhu et al. [30], who point out that PLA and PBAT offer favorable

degradability and high transparency, making them attractive for industrial applications. However, their

low mechanical strength necessitates recent efforts to improve the composite properties for packaging

applications.

Subsequently, the effect of "thickness" and "formulation" factors on puncture resistance was re-evaluated.

The optimal Box-Cox transformation parameter was λ = 1, so no transformation was applied to the

response variable.

Table 9 presents the ANOVA results for film puncture resistance. The data confirms that both thickness

and formulation are statistically significant factors (p ≤ 0.05) affecting this mechanical property, the only

property where both factors demonstrated relevant effects.

Table 9. Analysis of variance for punching resistance.

Source

SC Adjustment

MC Adjustment

F value

P-value

Biodegradable

0.5166

0.51663

9.75

0.002

Caliber

20.1195

2.87421

54.26

0.000

Error

241

12.7665

0.05297

Total

249

33.2135

Residual analysis confirmed compliance with ANOVA assumptions, including normality,

homoscedasticity, and error independence. Consequently, a post hoc multiple comparisons test was

performed using Fisher's method (LSD), the results of which are presented in Table 10.

Table 10. Fisher's LDS method and 95% confidence (punching strength).

Caliber (ga)

Media (N)

Group

300

3.26423

220

2.64553

210

2.38496

200

2.35424

190

2.31459

170

2.16301

180

100

2.15539

175

1.98779

The results show a direct relationship between film thickness and puncture resistance: greater thickness

requires higher force to fracture the material. The thinnest films exhibited the lowest resistance, with a

statistically significant difference between 170 ga (2.16 N) and 300 ga (3.26 N) thickness. This behavior

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

is primarily attributed to extruded film thickness, confirming the positive association observed in the

Pearson correlation analysis.

Furthermore, according to the means presented in Table 11, the biodegradable formulation demonstrated

superior puncture resistance compared to the non-biodegradable formulation, suggesting enhanced

mechanical performance of additive-containing materials under the evaluated conditions.

Table 11. Average punching resistance by formulation.

Biodegradable

Punching resistance (N)

Deviation standard

Yes

2.29

0.32

2.25

0.28

Furthermore, the effect of thickness and formulation factors on impact resistance was evaluated. To satisfy

ANOVA assumptions, a Box-Cox transformation with parameter λ = –1 was applied to the response

variable. Subsequently, variance analysis was conducted using the transformed variable.

Table 12 presents the ANOVA results, demonstrating that both thickness and formulation are statistically

significant factors (p ≤ 0.05) for the film's impact resistance. Residual analysis confirmed compliance with

ANOVA assumptions, including normality, homoscedasticity, and independence. Consequently, a post hoc

multiple comparisons test was performed using Fisher's method (LSD), the results of which are shown in

Table 13.

Table 12. Analysis of variance for response impact resistance transformation.

Source

SC Adjustment

MC Adjustment

F value

P-value

Biodegradable

0.000002

5.80

0.017

Caliber

0.000071

0.000010

38.51

0.000

Error

241

0.000064

0.0000000

Total

249

0.000136

Table 13. Fisher's LDS method and a 95% confidence (impact resistance).

Caliber (ga)

Media (g)

Group

300

456.635

220

281.081

210

271.896

175

248.846

200

241.416

190

239.725

180

100

229.813

170

216.818

The data demonstrates a positive relationship between film thickness and impact resistance: greater

thickness requires higher force to puncture the material. Thinner films recorded the lowest impact

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

resistance values, with a statistically significant difference between 170 ga (216.82 g) and 300 ga (3456.64

g) thicknesses.

This behavior is primarily attributed to extruded film thickness, confirmed by correlation analysis showing

a positive association between these variables. Therefore, increased thickness systematically enhances the

material's impact resistance capacity.

Additionally, according to the means reported in Table 14, the biodegradable formulation showed lower

impact resistance compared to the non-biodegradable formulation, suggesting a trade-off between

biodegradability and performance for this mechanical property.

Table 14. Average impact resistance by formulation.

Biodegradable

Impact resistance (g)

Deviation standard

Yes

259.68

56.65

268.10

48.07

When contrasting this study's results with those reported by Chai [10]—who documented average impact

resistance values of 456.64 g for 300 ga films and 775 g for 100 ga films—it is observed that the film

thickness analyzed here is approximately three times greater, correlating with superior resistance in free-

fall impact testing.

This difference can be attributed to varying experimental conditions and material types. In this study,

processing speed was 120 rpm, and the extrusion temperature range was maintained between 180 and 195

°C. As evidenced by Gálvez et al. [8], processing speed is a critical factor for extruded films' mechanical

properties.

Furthermore, experimental conditions—including applicable standards—directly influence results. This

study employed the Mexican standard NMX-E-099-CNCP-2014, which specifies a free-fall drop height

of 660 mm for the test dart. Variations in this height modify the applied potential energy upon impact,

affecting measured resistance.

Additionally, material types decisively affect mechanical properties. Films with biodegradable additives

exhibited lower impact and puncture resistance compared to non-biodegradable formulations, suggesting

a trade-off between biodegradability and material robustness.

3.2 Property regression models mechanics based in decision trees.

To describe and predict the mechanical properties of films based on variables such as thickness, thickness

deviation, extrusion temperature, and formulation, the M5P algorithm was applied. This method constructs

decision trees using the weighted reduction of standard deviation as the node-splitting criterion, fitting

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

linear regression models at terminal leaves. This constructs a decision tree where the splits are determined

by the weighted reduction in the standard deviation of the target variable. A linear regression is fitted to

each terminal leaf, allowing the tree segmentation to be combined with the precision of linear models. The

weighting expression for node splitting is shown in [31]:



󰇛󰇜󰇯󰇛󰇜  



󰇝󰇞󰇰 󰇛󰇜

As to:



 Proportion of examples without missing values for the evaluated attribute.

󰇛󰇜 Correction factor that penalizes attributes with many values.

󰇛󰇜 Standard deviation of the class values at the current node T.





󰇝󰇞 Weighed sum of the standard deviations of the left (L) and right (R) sub nodes.

To apply the M5P algorithm for modeling the mechanical properties of films based on operating variables,

WEKA software version 3.9.6 was used. To obtain models with optimal fit, various transformations were

evaluated for the response variables, selecting those exhibiting the highest correlation coefficient. The

identified optimal transformations are presented in Table 15.

Table 15. Transformations with best fit from the correlation coefficient.

Property mechanics

Transformation

Tensile strength

Quadratic

Elongation

Logarithmic

Tear resistance

Quadratic

Punching resistance

Quadratic

Furthermore, since the primary hyperparameter of the M5P algorithm corresponds to the minimum

number of instances required to fit the model at each leaf, different criteria were evaluated by varying this

parameter. Specifically, varying minimum instance quantities per leaf were analyzed for linear model

estimation. Table 16 presents the correlation coefficients obtained for each mechanical property across

different minimum instance per leaf values.

Table 16. Correlation coefficients varying the minimums instances per sheet (R).

Instances

per sheet

Tensile

strength

Elongation

Tear

resistance

Punching

resistance

0.6557

0.4662

0.8047

0.5746

0.6557

0.4662

0.8047

0.5746

0.6533

0.4999

0.7825

0.5811

0.6547

0.4943

0.7827

0.5811

0.6713

0.5266

0.7925

0.5811

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

0.6680

0.4784

0.7929

0.5811

0.6652

0.4849

0.7817

0.5771

As observed, the mechanical property best described by the input data (thickness, thickness deviation,

extrusion temperature, and formulation) is tearing resistance, with a correlation coefficient of 0.8047

obtained when training a model using 4 minimum instances per leaf. Conversely, tensile strength exhibited

a correlation coefficient of 0.6713 with 10 minimum instances per leaf. It is worth mentioning that the

mathematical models presented below are expressed according to their respective linearization used to

apply the algorithm.

This result can be compared to that reported by [32], where a decision tree algorithm applied to

experimental tensile strength data for high-density polyethylene (HDPE) films achieved a correlation

coefficient of 0.94 and a mean absolute error of 0.04 with 14 minimum instances per leaf. The lower

correlation in this study may be explained by different predictor variables: while [32] used only operating

temperatures and mechanical properties, the present work also considered material-specific variables like

density and melt flow index—whose inclusion could improve model fit as noted in [14] .

For elongation and puncture resistance, correlation coefficients of 0.5266 and 0.5811 were obtained,

respectively. These values indicate that while the models partially explain the variation in these properties,

other unconsidered variables affect their behavior, and their inclusion could enhance predictive capability.

The tearing resistance model is presented in Equation (2). Its utility lies in evaluating whether a process—

under established conditions and with specific thickness—meets customer specifications. This property is

particularly relevant given the high correlation (0.8047) achieved in the final model.

Tearing r. = 0.000011 * Cr + 0.000008 * Te - 0.307354 (2)

As to:

Cr = Actual caliber of the film.

Te = Extrusion temperature.

The model obtained for tearing resistance consists of a single rule, indicating a linear fit between this

property and the average extrusion temperature. The mean absolute error was 0.0482, equivalent to 0.2195

kgf, considering the tearing resistance data underwent quadratic transformation.

For tensile strength, a quadratic transformation was similarly applied, and the optimal model is shown in

Equations (3) and (4), using actual thickness as the splitting criterion.

Cr ≤ 31289.4165

Tensile r. = 0.011109*Cr+515.45731 (3)

Cr > 31289.4165

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

Tensile r. = 101.227387 * B + 0.014527 * Cr + 556.00864 (4)

As to:

B = Biodegradable (No=0, Yes=1).

Cr = Actual caliber of the film.

As observed, the obtained model consists of two rules, where the actual thickness selects one of the two

presented linear models. For thicknesses greater than 31,289.4165 ga, formulation becomes particularly

relevant, as the second linear model includes a term corresponding to formulation—assigned a value of 1

for non-biodegradable formulations and 0 for biodegradable formulations.

This behavior is explained by the P–Life biodegradable additive, incorporated at 1% mass proportion. Its

primary function is to break long-chain polyethylene polymers into shorter molecules (oligomers) through

an oxidation process typically activated by ultraviolet light, heat, and oxygen.

Regarding model performance, the mean absolute error was 154.4205 (transformed), equivalent to

12.4266 MPa, considering a quadratically transformed model. The models for elongation and puncture

resistance are presented in Equations (5) and (6).

Elongation = 0.475258 * Cr - 0.121955 * D + 4.331964 (5)

As to:

Cr = Actual caliber of the film.

D = Film miscaliber.

Punching r. = 0.000094 * Cr + 0.00008 * Te - 0.984355 (6)

As to:

Cr = Actual caliber of the film.

Te = Extrusion temperature.

The correlation coefficients for elongation and puncture resistance were 0.5266 and 0.5811, respectively.

Similarly, the mean absolute error (MAE) was 0.0666 for elongation and 0.7389 for puncture resistance,

equivalent to a percentage error of 1.069% for elongation and an absolute error of 0.860 N for puncture

resistance.

For these two mechanical properties, it is necessary to include additional variables to increase the

explained variance percentage, improve the correlation coefficient, and reduce model prediction errors.

Notably, logarithmic transformations were applied for elongation and quadratic transformations for

puncture resistance.

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

The decision tree-based regression model was evaluated with 10 independent samples excluded from the

training set. The test results, including the error associated with each mechanical property, are presented

in Table 17.

Table 17. Testing the decision tree model.

Proof

Error in tensile

strength

Error in

elongation

Error in tear

resistance

Error in punching

resistance

1.210

5.178

2.070

3.607

1.795

2.622

0.695

2.902

3.788

12.602

0.196

6.076

4.277

13.877

3.701

0.583

6.903

10.626

0.543

9.303

0.108

13.766

4.444

11.259

3.184

0.035

2.959

3.163

22.101

6.289

9.411

1.291

9.446

6.849

13.211

14.135

1.878

1.956

9.817

6.250

Based on the data from Table 17, the average error for each mechanical property was calculated to quantify

individual errors. The resulting values are presented in Table 18.

Table 18. Average error for each property mechanics.

Error in tensile

strength (%)

Error in

elongation (%)

Error in tear

resistance (%)

Error in punching

strength (%)

5.469

7.380

4.705

5.857

Finally, based on the data presented in Table 18, the overall average error was calculated, yielding a value

of 5.853%. This quantifies the decision tree-based regression model's accuracy for predicting mechanical

properties at 94.147%. This accuracy level is validated according to the ASME V&V 20-2021 standard—

primarily used to quantify model accuracy against experimental data—which deems a model acceptable

when achieving overall accuracy ≥90% [33, 34].

These results demonstrate the efficacy of machine learning algorithms for modeling mechanical properties

in polymer manufacturing processes. Specifically, nonlinear decision tree-based methods are essential for

capturing complex multivariable relationships and optimizing production. Moreover, the ability to identify

key data patterns—such as formulation and processing parameters—reinforces these models' industrial

utility by reducing reliance on empirical testing and enhancing quality control.

4 Conclusion

This study evaluated the impact of the biodegradable additive P-Life on the mechanical properties of low-density

polyethylene (LDPE) films for ice bags and developed a predictive model based on artificial intelligence. The

ANOVA results identified film thickness as the most influential factor for tensile strength, elongation, and tear

resistance, showing statistically significant differences (p ≤ 0.05). In contrast, the formulation (with or without the

biodegradable additive) did not exhibit a significant effect on these same properties. However, a discernible

influence of P-Life was observed on puncture and impact resistance. This selective effect can be attributed to the

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

additive's mechanism of action, which is designed to initiate degradation through an oxidation process. By acting

as localized sites for stress concentration, these modifiers alter the material's ability to absorb energy under point

loads or impact, thereby modifying these specific properties without significantly affecting the behavior under

tensile or tearing stresses.

The M5P machine learning algorithm proved to be an effective tool, predicting mechanical properties with an

accuracy of 94.147%. This validates its utility for optimizing the extrusion process and reducing reliance on

empirical testing methods. Collectively, this work confirms that while thickness is the critical parameter for most

mechanical properties, the incorporation of the P-Life additive introduces specific changes in the material's

performance. Furthermore, it corroborates the feasibility of implementing artificial intelligence models for quality

control and the development of more sustainable flexible packaging.

5. Authorship acknowledgements

Gilberto Alarcón Aguilar: Conceptualization; Ideas; Methodology; Research; Writing; Original Draft. Alam Josué

Reyes López: Conceptualization; Data Analysis. Frixia Galán-Méndez: Conceptualization; Methodology; Formal

Analysis; Research; Writing; Revision and Editing; Project Management.

References

[1] I. Santos Silva et al., "Malt bagasse in extruded wheat flour and PBAT biodegradable films," Ind. Crops Prod.,

vol. 220, p. 119274, Nov. 2024, doi: 10.1016/j.indcrop.2024.119274.

[2] A. Martínez-Camacho et al., "Extruded films of blended chitosan, low density polyethylene and ethylene acrylic

acid," Carbohydr. Polym., vol. 91, pp. 666–674, Aug. 2012, doi: 10.1016/j.carbpol.2012.08.076.

[3] H. Najahi et al., "Plastic pollution in food packaging systems: Impact on human health, socioeconomic

considerations and regulatory framework," Elsevier B.V., May 2025, doi: 10.1016/j.hazadv.2025.100667.

[4] B. Ucpinar, F. Ugur, and A. Aytac, "Active packaging films based on poly(butylene succinate) films reinforced

with alkaline halloysite nanotubes: Production, properties, and fruit packaging applications," Appl. Clay Sci., vol.

256, p. 107517, Jul. 2024, doi: 10.1016/j.clay.2024.107517.

[5] L. Yi et al., "Simple but efficient preparation of high-strength, heat-resistant, and high-barrier PLA/TOBC plastic

packaging containers via Pickering emulsion," Compos. Commun., vol. 56, p. 102420, Jun. 2025, doi:

10.1016/j.coco.2025.102420.

[6] P-Life American, "Aditivos plásticos biodegradables. Plásticos biodegradables, una realidad,"

plifeamerican.com. https://plifeamerican.com (accessed May 5, 2025).

[7] T. Zhang, W. Han, C. Zhang, and Y. Weng, "Effect of chain extender and light stabilizer on the weathering

resistance of PBAT/PLA blend films prepared by extrusion blowing," Polym. Degrad. Stabil., vol. 183, p. 109455,

Jan. 2021, doi: 10.1016/j.polymdegradstab.2020.109455.

[8] J. Gálvez et al., "Effect of extrusion screw speed and plasticizer proportions on the rheological, thermal,

mechanical, morphological and superficial properties of PLA," Polymers, vol. 12, no. 9, p. 2111, Sep. 2020, doi:

10.3390/polym12092111.

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

[9] H. Wu et al., "Mechanical activation-enhanced metal-organic coordination strategy to Fabricate high-

performance starch/polyvinyl alcohol films by extrusion blowing," Carbohydr. Polym., vol. 333, p. 121982, Feb.

2024, doi: 10.1016/j.carbpol.2024.121982.

[10] C. K. Chai, "Rheological studies of molecular effect and processing conditions on blown film property of

polyethylenes," Polymer, vol. 267, p. 125668, Feb. 2023, doi: 10.1016/j.polymer.2022.125668.

[11] B. E. Itabana, A. K. Pal, A. K. Mohanty, and M. Misra, "Biodegradable blown film composite from

poly(butylene adipate-co-terephthalate) and talc: Effect of uniaxial stretching on mechanical and barrier properties,"

Food Packag. Shelf Life, vol. 39, p. 101147, Nov. 2023, doi: 10.1016/j.fpsl.2023.101147.

[12] J. Park, K.-Y. Kim, and R. Sohmshetty, "A prediction modeling framework: Toward integration of noisy

manufacturing data and product design," ASME, 2015. [Online]. Available: http://www.asme.org/about-

asme/terms-of-use

[13] H. A. Dahish and A. D. Almutairi, "Compressive strength prediction models for concrete containing nano

materials and exposed to elevated temperatures," Results Eng., vol. 25, p. 103975, Mar. 2025, doi:

10.1016/j.rineng.2025.103975.

[14] S. Altarazi, R. Allaf, and F. Alhindawi, "Machine learning models for predicting and classifying the tensile

strength of polymeric films fabricated via different production processes," Materials, vol. 12, no. 9, p. 1475, 2019,

doi: 10.3390/ma12091475.

[15] S. K. Dang and Singh, "Predicting tensile-shear strength of nugget using M5P model tree and random forest:

An analysis," Comput. Ind., vol. 124, p. 103345, Jan. 2021, doi: 10.1016/j.compind.2020.103345.

[16] C. Herriott and A. D. Spear, "Predicting microstructure-dependent mechanical properties in additively

manufactured metals with machine- and deep-learning methods," Comput. Mater. Sci., vol. 175, p. 109599, Apr.

2020, doi: 10.1016/j.commatsci.2020.109599.

[17] M. Abdallah et al., "The Machine-Learning-Based Prediction of the Punching Shear Capacity of Reinforced

Concrete Flat Slabs: An Advanced M5P Model Tree Approach," Appl. Sci., vol. 13, p. 8325, Jul. 2023, doi:

10.3390/app13148325.

[18] C. Machello et al., "Tree-based machine learning approach to modelling tensile strength retention of Fibre

Reinforced Polymer composites exposed to elevated temperatures," Compos. Part B Eng., vol. 270, p. 111132, Nov.

2023, doi: 10.1016/j.compositesb.2023.111132.

[19] Y. Boon, S. Joshi, S. Bhudolia, and G. Gohel, "Recent Advances on the Design Automation for Performance-

Optimized Fiber Reinforced Polymer Composite Components," Compos. Sci., vol. 4, p. 60, May 2020, doi:

10.3390/jcs4020061.

[20] A. Mohammed et al., "Modeling the Impact of Liquid Polymers on Concrete Stability in Terms of a Slump and

Compressive Strength," Appl. Sci., vol. 13, p. 1208, Jan. 2023, doi: 10.3390/app13021208.

[21] E. Temizhan, H. Mirtagioglu, and M. Mendes, "Which Correlation Coefficient Should Be Used for

Investigating Relations between Quantitative Variables?," Am. Sci. Res. J. Eng. Technol. Sci., vol. 61, p. 5524, Dec.

2021. [Online]. Available: https://www.researchgate.net/publication/359579944

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

[22] A. Atkinson, M. Riani, and A. Corbellini, "The Box–Cox Transformation: Review and Extensions," Stat. Sci.,

vol. 36, pp. 239–255, 2021, doi: 10.1214/20-STS778.

[23] J. Lu, B. Nguyen, and J. Powers, "Mechanical properties of 3 hydrophilic addition silicone and polyether

elastomeric impression materials," J. Prosthet. Dent., vol. 467, p. 6516, Mar. 2004, doi:

10.1016/j.prosdent.2004.05.016.

[24] N. Mallegni, T. V. Phuong, M. B. Coltelli, P. Cinelli, and A. Lazzeri, "Poly(lactic acid) (PLA) based tear

resistant and biodegradable flexible films by blown film extrusion," Materials, vol. 11, no. 1, p. 148, Jan. 2018, doi:

10.3390/ma11010148.

[25] L. Aliotta et al., "Tearing fracture of poly(lactic acid) (PLA)/poly(butylene succinate-co-adipate) (PBSA) cast

extruded films: Effect of the PBSA content," Eng. Fract. Mech., vol. 289, p. 109450, Sep. 2023, doi:

10.1016/j.engfracmech.2023.109450.

[26] L. Gómez-Bachar, M. Vilcovsky, P. González-Seligra, and L. Famá, "Effects of PVA and yerba mate extract

on extruded films of carboxymethyl cassava starch/PVA blends for antioxidant and mechanically resistant," Int. J.

Biol. Macromol., vol. 268, p. 131464, Apr. 2024, doi: 10.1016/j.ijbiomac.2024.131464.

[27] S. Kim, S. Kang, Y. Kim, and M. Park, "Industrial-scale blown active packaging film with essential oils:

Properties, dual-functional performance, and box packaging application of instant noodles," Food Packag. Shelf

Life, vol. 43, p. 101276, Apr. 2024, doi: 10.1016/j.fpsl.2024.101276.

[28] X. Zhang, S. Ajji, and M. Huneault, "Oriented structure and anisotropy properties of polymer blown films:

HDPE, LLDPE and LDPE," Polymer, vol. 45, pp. 217–229, Oct. 2003, doi: 10.1016/j.polymer.2003.10.057.

[29] W. Zhou, D. Zha, X. Zhang, J. Xu, B. Gou, and Y. Huang, "Ordered long-period stacking ordered phase..."

[Incompleta en el texto original].

[30] B. Zhu et al., "Enhancing the mechanical properties of polylactic acid (PLA) composite films using Pueraria

lobata root microcrystalline cellulose," Int. J. Biol. Macromol., vol. 279, p. 135579, Nov. 2024, doi:

10.1016/j.ijbiomac.2024.135579.

[31] I. Witten, "Inducing model trees for continuous classes," 1997. [Online]. Available:

https://www.researchgate.net/publication/2737587

[32] S. Altarazi, M. Ammouri, and A. Hijazi, "Artificial neural network modeling to evaluate polyvinylchloride

composites’ properties," Comput. Mater. Sci., vol. 153, pp. 1–9, Oct. 2018, doi: 10.1016/j.commatsci.2018.06.003.

[34] K. Balkey, A. Domenic, A. Guzmán, and S. Weinman, "Ejemplos de uso de códigos y normas para los

estudiantes de ingeniería mecánica y otros campos," 2021. [Online]. Available: http://www.asme.org/about-asme

[35] A. Gaspar-Cunha, J. A. Covas, and J. Sikora, "Optimization of polymer processing: A review (Part II - Molding

technologies)," Materials, vol. 15, no. 3, p. 1138, Feb. 2022, doi: 10.3390/ma15031138.

Revista de Ciencias Tecnológicas (RECIT). Volumen 9 (2): e420.

ISSN: 2594-1925

Este texto está protegido por una licencia Creative Commons 4.0.

Usted es libre para compartir —copiar y redistribuir el material en cualquier medio o formato — y adaptar el documento —

remezclar, transformar y crear a partir del material— para cualquier propósito, incluso para fines comerciales, siempre que

cumpla la condición de:

Atribución: Usted debe dar crédito a la obra original de manera adecuada, proporcionar un enlace a la licencia, e indicar si se

han realizado cambios. Puede hacerlo en cualquier forma razonable, pero no de forma tal que sugiera que tiene el apoyo del

licenciante o lo recibe por el uso que hace de la obra.

Resumen de licencia - Texto completo de la licencia