Page 1032
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
Machine Learning-RSM Hybridized Evaluation of the Kinetics and
Thermodynamics of Mild Steel Corrosion Inhibition Using Lagenaria
Breviflora Extract
Nnorom Obinichi
1
, Ifeanyi Uchegbulam
2
, Opuwil Samuel Chimenem
3
1
Department of Mechanical Engineering, Faculty of Engineering, University of Port Harcourt, Choba,
P.M.B., 5323, Nigeria
2
Production Technology, School of Science Laboratory Technology, University of Port Harcourt, Choba,
P.M.B., 5323, Nigeria.
3
Industrial and Manufacturing Engineering, FAMU-FSU College of Engineering, Tallahassee, FL
32310, United States
DOI:
https://doi.org/10.51583/IJLTEMAS.2026.150300089
Received: 22 March 2026; Accepted: 27 March 2026; Published: 17 April 2026
ABSTRACT
The inhibition of mild steel corrosion in dilute hydrochloric acid (1 M HCl) by the xylene extract of Lagenaria
breviflora (XEL-B) was studied using a Central Composite Design (CCD) structured Response Surface
Methodology (RSM). A statistically optimised 20-run experimental matrix was employed to evaluate the
simultaneous effects of inhibitor concentration, immersion time and temperature on mass loss, corrosion rate
(Rc), inhibitor efficiency (IE), and surface coverage (θ). The fitted quadratic response surface model was highly
significant (F = 55.81, p < 0.0001) with a non-significant lack of fit (p = 0.2730), confirming adequate model
predictability across the experimental domain. Inhibitor efficiency ranged from 28.68% at 31 ppm of inhibitor
concentration to 77.01% at 368 ppm, with inhibitor concentration identified as the dominant process variable
statistically validated (F = 423.40, p < 0.0001) cosnsistent with 4D response surface analysis, and a 500-tree
Random Forest ensemble machine learning model (factor importance: IE = 77.54%, Rc = 74.20%). Adsorption
of XEL-B on mild steel conformed to the Langmuir monolayer isotherm (R² = 0.9950), equilibrium adsorption
constant K
ads
of 11.3649Lg⁻¹ and standard Gibbs free energy of adsorption ΔG°ads of −16.24 kJ/mol, confirming
spontaneous, thermodynamically favourable adsorption with a mixed physisorptive–chemisorptive mechanism.
These integrated experimental-computational results established XEL-B as a potential green corrosion inhibitor
for mild steel in dilute acidic environments relevant to oilfield and industrial acid-treatment operations.
Keywords: Lagenaria breviflora; corrosion inhibition; steel; RSM; Langmuir isotherm; Random Forest;
machine learning
INTRODUCTION
Acidizing, also referred to as acid treatment, is a well-established and highly effective technique for enhancing
oil and gas well productivity. Hydrochloric acid-based solutions are widely employed during this process to
improve reservoir permeability and stimulate hydrocarbon flow. With over a century of application, acidizing
remains one of the earliest and most reliable well stimulation methods, predating techniques such as hydraulic
fracturing, (Iroha and Akaranta, 2020). Meanwhile, the HCL acid content of acidizing solutions exposes metallic
components to severe corrosion-induced degradation. Oil pipelines, tubing, and casings are the most susceptible
to these acidizing well stimulation process (Ituen et al , 2021). With corrosion related damages costing up to 2.5
trillion USD per year covering approximately 3.4% of the global Gross Domestic Product (GDP) (Mohammad
and Jafar, 2020; Kania, 2023; Adesina et al., 2025; Zakeri et al., 2022). Due to this huge economic loss, corrosion
control using corrosion inhibitors has been a rewarding approach to preserving the durability of industrial assets,
(Bandeira et al., 2025). Plant extracts have emerged as highly effective green corrosion inhibitors due to their
Page 1033
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
availability, non-toxicity, biodegradability, and renewable nature (Otaibi et al., 2021). Rich in phytochemicals
such as alkaloids, tannins, flavonoids, and polyphenols, these natural compounds adsorb onto metal surfaces to
form protective films that suppress electrochemical corrosion by influencing anodic and cathodic reactions
through donor–acceptor interactions (Barbu et al., 2025). Historically, the use of plant-based inhibitors dates
back to the 1930s and has since evolved with increasing recognition of their efficiency and environmental
benefits, particularly as sustainable alternatives to conventional inhibitors, (Kumari and Lavanya, 2022). When
introduced in small quantities into corrosive media, these inhibitors function by forming a protective barrier that
reduces metal degradation (Alao et al., 2022). It has been reported that inhibitor concentration, immersion time
and temperature are determinant factors to inhibitor efficiency, (Alamiery et al., 2021; Al-Baghdadi et al., 2021).
Meanwhile, the feature importance of these factors has not been investigated within the corrosion inhibition
research landscape. The advent of machine learning and its algorithms has simplified feature ranking. Feature
importance techniques, as part of explainable artificial intelligence (XAI), enhance the interpretability of
machine learning models by enabling users to understand and trust model predictions (Cappelli et al., 2024). In
particular, Random Forest (RF) models provide inherent capability to quantify the relative importance of input
features, offering valuable insights into the underlying data relationships (Yuan et al., 2023). In this study, an
integration of the statistical novelty of response surface methodology and random forest as a machine learning
algorithm for feature importance ranking were deployed to investigate the inhibitive efficiency of Xylene extract
of Lagenaria Breviflora leaves as potential corrosion inhibitor for mild steel in acidized environments.
MATERIALS AND METHODS
Plant Material and Xylene Extraction
Mature leaves of Lagenaria Breviflora were collected from a botanical garden adjacent to Nigerian Television
Authority Port Harcourt Nigeria and was authenticated by a certified taxonomist. The leaves were surface-
washed with distilled water, air-dried at ambient temperature (298 ± 2 K) for 21 days prior to an oven-drying at
323 K for 48 h to a constant mass. The dried leaves were pulverised to a fine powder (<250 μm, 60-mesh sieve)
using a stainless-steel laboratory mill. Then, 100 g of dried powder was loaded into a Soxhlet extraction thimble
and extracted continuously for 8 h at reflux with 500 mL of analytical-grade xylene (Sigma-Aldrich, ≥98.5%;
bp 138–144 °C). The resulting extract was vacuum-filtered through Whatman No. 1 filter paper and the solvent
removed under reduced pressure using a rotary evaporator (333 K, 200 mbar) with all laboratory work carried
out at the Africa Center for Excellence center for oilfield research at University of Port Harcourt Nigeria.
Mild Steel Coupon Preparation
Mild steel coupons were machined into rectangular specimens providing a nominal exposed surface area of 10.0
cm². Elemental composition was estimated as wt%: C 0.18, Si 0.21, Mn 0.68, P 0.012, S 0.008, Fe balance using
an Oxford Instrument X-Met 7000 XRF Spectrometer, (Turret Engineering Services Ltd, Port Harcourt Nigeria).
Surface preparation involved mechanical abrasion using SiC metallographic papers of different grits: 240, 400,
600, 800, and 1200. The shiny surfaces were degreased in acetone, rinsed with distilled water and oven-dried.
Initial and final masses were recorded to ±0.0001g precision on a calibrated analytical balance.
Corrosive Medium
The corrosive medium was 1 M hydrochloric acid (HCl), prepared by diluting concentrated analytical-grade HCl
(37% w/w, Sigma-Aldrich) with double-distilled water. The acid concentration was verified by titrimetric
standardisation against primary-standard anhydrous Na
2
CO
3
using methyl orange indicator. Fresh acid solutions
were prepared for each experimental run.
Central Composite Design (CCD) Experimental Framework
A three-factor, five-level Central Composite Design (CCD) was constructed using Design-Expert® v13.0 (Stat-
Ease Inc., Minneapolis, MN, USA). The CCD incorporated 2³ = 8 factorial points, 6 axial (star) points at ±α =
±1.682 (rotatability condition), and 6 centre point replicates to generate a 20-run design matrix capable of fitting
a complete second-order quadratic response surface model. The independent variables and their coded levels
Page 1034
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
were: Factor A being inhibitor concentration (31, 100, 200, 300, 368 ppm; coded: −1.682, −1, 0, +1, +1.682);
Factor B which was immersion time (1, 1.5, 2, 3, 4 h; coded: −1.682, −1, 0, +1, +1.682) and Factor C being
temperature (294, 298, 303, 308, 311 K; coded: −1.682, −1, 0, +1, +1.682). The response variables were mass
loss (ΔW, mg), corrosion rate (Rc, mg cm⁻² h⁻¹), inhibitor efficiency (IE, %), and surface coverage (θ).
Mass Loss Procedure and Corrosion Parameter Calculations
Pre-weighed coupons were immersed in 100 mL of test solution in sealed 250 mL polypropylene beakers
maintained at the designated temperature in a thermostatically controlled water bath (Grant TC120, ±0.2 K).
After immersion, coupons were retrieved, transferred immediately to inhibitor-free acid solution to arrest
corrosion, scrubbed gently under running distilled water, rinsed with acetone, dried, and reweighed. All runs
were performed in triplicate; mean values are reported (RSD < 3.5%). Corrosion parameters were derived as:
2.1
Where = mass loss (mg)
W
I
= initial coupon mass before immersion (mg)
W
F
= final coupon mass after corrosion and cleaning (mg)
Also,
2.2
Where
= Corrosion rate (mg cm⁻² h⁻¹)
= mass loss (mg)
A = exposed surface area (10.0 cm
2
)
T = immersion time (h)
And
2.3
Where
= corrosion rate in uninhibited (blank) solution (mg cm⁻² h⁻¹)
= corrosion rate in inhibited solution (mg cm⁻² h⁻¹)
While
2.4
Where = fractional surface coverage (dimensionless, 0 ≤ θ ≤ 1)
Adsorption Isotherm Analysis
The adsorption behaviour of Lagenaria Breviflora leaf extracts on mild steel was modelled using the linearised
Langmuir adsorption isotherm, which postulates monolayer adsorption on a finite number of equivalents,
energetically homogeneous, and mutually non-interacting surface sites. Surface coverage values (θ) derived from
the experimental IE data were plotted against the corresponding inhibitor concentrations according to the
linearized Langmuir expression:
2.5
Where
= inhibitor concentration (g L⁻¹)
Page 1035
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
= fractional surface coverage (dimensionless, 0 ≤ θ ≤ 1)
= equilibrium adsorption constant for the adsorption process (L g⁻¹)
The standard Gibbs free energy of adsorption (
) is related to the equilibrium adsorption constant
as:
2.6
Where
= equilibrium adsorption constant (L g⁻¹)
= standard Gibbs free energy of adsorption (J)
R = universal gas constant (8.314 J mol⁻¹ K⁻¹)
T = absolute temperature (K)
55.5 = molar concentration of pure water (mol L⁻¹)
On reordering equation 2.6:
2.7
A three-dimensional Langmuir adsorption surface was additionally constructed to model predicted adsorption
capacity (q
sim
, mg/g) as a simultaneous function of equilibrium concentration (Ce, mg/L) and temperature (T,
°C):
2.8
Where
= predicted adsorption capacity (mg g⁻¹)
KL = temperature-dependent Langmuir adsorption constant (L mg⁻¹)
= maximum monolayer adsorption capacity, temperature-dependent (mg g⁻¹)
Ce = equilibrium inhibitor concentration in solution (mg L⁻¹)
Thermodynamic and Activation Parameter Determination
The activation energy of corrosion in blank and inhibited solutions were determined from the linearised
Arrhenius equation:
2.9
Where A = Arrhenius pre-exponential factor
E
a
= apparent activation energy of corrosion (kJ mol⁻¹)
R = universal gas constant (8.314 J mol⁻¹ K⁻¹)
T = absolute temperature (K)
Enthalpy (ΔH*) and entropy (ΔS*) of activation were evaluated from the Eyring–Polanyi transition state
equation:
Page 1036
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
2.10
Where N = Avogadro’s number (6.022 × 10²³ mol⁻¹)
h = Planck’s constant (6.626 × 10⁻³⁴ J s)
= enthalpy of activation (kJ mol⁻¹)
= entropy of activation (J mol⁻¹ K⁻¹)
Linear plots of ln(Rc/T) versus 1/T yielded ΔH* from the slope and ΔS* from the intercept.
Random Forest Machine Learning Feature Importance Analysis
To provide a non-parametric, data-driven ranking of the three process variables — inhibitor concentration (A),
immersion time (B), and temperature (C) — with respect to their influence on IE and Rc, a Random Forest (RF)
ensemble regression algorithm was implemented in Python 3.11 using scikit-learn v1.4. Random Forest
constructs an ensemble of n = 500 decorrelated decision trees, each trained on a bootstrap-sampled subset of the
training observations; the mean decrease in node impurity (MDI, Gini criterion) across all trees provides the
feature importance score, normalised to sum to unity. Hyperparameters: n_estimators = 500, max_features = √p
(p = 3), random_state = 42. Permutation importance (100 repetitions, random_state = 42) and Gradient Boosting
Regression (GBR; n_estimators = 200, learning_rate = 0.05) were applied as independent cross-validation
procedures. All importance scores are reported as percentages of total explained variance.
RESULTS AND DISCUSSION
Statistical modeling of process parameters
The 20-run CCD experimental matrix was used to measure and compute corrosion response data. The
uninhibited (blank) corrosion rates ranged from 3.9583 mg cm⁻² h⁻¹ at 294 K to 4.8404 mg cm⁻² h⁻¹ at 311 K,
conforming to Arrhenius kinetics and confirming the temperature sensitivity of acid corrosion of mild steel.
Introduction of Xylene extracts of Lagenaria Breviflora leaves across all 20 CCD runs produced consistent,
concentration-dependent suppression of both mass loss and corrosion rate.
Figure 3.1 is the ANOVA for the fitted quadratic response surface model. It shows that the model is highly
statistically significant (F-value = 55.81) with P< 0.0001 which was less than the confidence interval of 0.05.
Also, figure 3.1 shows the significance of the model terms. Term A (Inhibitor Concentration has p < 0.0001 and
is the most significant individual factor while terms A and B (Immersion time and temperature) are not
statistically significant, confirming Random Forest findings (77.54% importance). significant model terms were
A (Inhibitor concentration), AB, BC and A
2
. Meanwhile, the interactive effect of inhibitor concentration with
both immersion time and temperature were significant as well as the quadratic interaction of the inhibitor
concentration. Remarkably, the model is fit to predict the inhibitor efficiency as it fits the data adequately with
no systematic misfit. As the lack of fit scored an F-value of 1.77, it implies that the lack of fit is not significant
and that there was 27.30% chance that a lack of fit F-value this large could occur due to noise. confirming the
model.
Page 1037
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
Figure 3.1: ANOVA for the quadratic model
Also, equation 3.1 shows the predictive model: IE (%) = 3875.1116 - 0.2223x + 123.8350y - 25.7996z +
0.0308xy + 0.0015xz - 0.4150yz -0.0004x
2
- 0.9339y
2
+ 0.0433z
2
Where x, y and z are the inhibitor concentration, time of immersion and Temperature respectively.
Model Diagnostic Plots
Figure 3.2 shows four standard CCD model adequacy diagnostics. Figure 3.2a shows that the points cluster
reasonably along the 45° diagonal, confirming acceptable model predictive accuracy. Figure 3.2 b shows the
Box-Cox Plot. Since the Current λ = 1and lies within the best lambda λ = 2.18, then there is no need for further
transformation as recommended. Hence, the raw IE response requires no power transformation and the linear
model is appropriate.
Page 1038
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
Figure 3.2: Model Diagnostic Plots (a) Predicted vs. Actual (b) Box-Cox Plot (c) Residuals vs. Predicted (d)
Normal Plot of Residuals
Likewise, figure 3.2 c shows that the residuals were randomly scattered within the ±4.14579 studentised limit
with no systematic pattern, confirming homoscedasticity (constant variance) as a key regression assumption.
Similarly, points followed the straight reference line closely in figure 3.2d thereby validating the statistical
assumptions of the quadratic CCD model.
3D Response Surface Plots
Furthermore, figure 3.3 shows four panels visualising how Inhibitor Efficiency (IE) responded to the three
process variables. Figure 3.3a shows that the IE rises steeply with increasing inhibitor concentration but is
relatively insensitive to immersion time. It can be seen that the surface was near-flat along the time axis visually
confirmed that inhibitor concentration is the dominant driver of IE.
Likewise, figure 3.3b shows that inhibitor concentration drove the steep gradient while temperature produced
minimal curvature. However, in figure 3.3c, the surface was almost perfectly flat (yellow/green plane),
confirming that neither temperature nor immersion time independently exerts strong influence on IE within the
studied range. Figure 3.3d shows the 3D Langmuir surface.
It matched the q_sim vs. Ce vs. T adsorption capacity surface showing monotonically increasing adsorption
capacity with both inhibitor concentration and temperature confirming endothermic Langmuir monolayer
behaviour.
Page 1039
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
Figure 3.3: 3D Response Surface Plots (a) IE vs. Concentration & Time (b) IE vs. Concentration & Temperature
(c) IE vs. Temperature & Time (d) 3D Langmuir surface
The three-dimensional Langmuir adsorption surface (Figure 3.3d) was constructed using equation 2.8 to
simultaneously visualise the predicted adsorption capacity q_sim (mg g⁻¹) as a function of equilibrium inhibitor
concentration Ce (mg L⁻¹) and temperature T (°C).
The surface demonstrated monotonically increasing q_sim with both Ce and T, a hallmark of endothermic
adsorption, reaching a maximum predicted value of ~477 mg g⁻¹ at Ce = 2000 mg L⁻¹ and T = 100 °C. The
pronounced curvature of the surface with respect to Ce with rapid initial rise flattening toward a saturation
plateau captures the inhibitor concentration-saturation characteristic of Langmuir monolayer kinetics, visually
Page 1040
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
corroborating the isotherm linearisation results. The three-dimensional representation provided an intuitive
mechanistic map of the inhibitor adsorption landscape across the practically relevant inhibitor concentration and
temperature space.
Thermodynamic and Activation Parameters
The activation energy of corrosion in the blank system was Ea = 43.8 kJ mol⁻¹, increasing substantially to 59.4
kJ mol⁻¹ in the presence of 200 ppm of the xylene extracts of Lagenaria Breviflora leaves resulted in a differential
of +15.6 kJ mol⁻¹. This elevation of activation energy in the inhibited system is the mechanistic signature of
physical adsorption-type inhibition as the extract molecules adsorbed onto the steel surface creating an energy
barrier that substantially retarded the electrochemical dissolution of iron.
The negative ΔG°ads values (−15.76 to −16.67 kJ mol⁻¹) confirmed a spontaneous and thermodynamically
favourable adsorption; values in this range are characteristic of mixed physisorption–chemisorption, consistent
with the multifunctional phytochemical composition of the xylene extract.
Table 3.1 Temperature-dependent corrosion parameters and thermodynamic quantities for mild steel
T (K)
IE (%)
Θ
Rc Blank
(mg cm⁻² h⁻¹)
Rc Inh.
(mg cm⁻² h⁻¹)
ΔG°ads
(kJ mol⁻¹)
Ea (blank)
(kJ mol⁻¹)
Ea (inh.)
(kJ mol⁻¹)
294
68.25
0.6825
3.9583
1.2568
−15.76
43.8
59.4
298
—
—
4.1853
—
−15.96
43.8
59.4
303
64.50
0.6450
4.5000
1.5976
−16.24
43.8
59.4
308
—
—
4.8434
—
−16.52
43.8
59.4
311
67.99
0.6799
4.9534
1.5856
−16.67
43.8
59.4
Langmuir Adsorption Isotherm and Three-Dimensional Adsorption Surface
The adsorption mechanism of the xylene extracts of Lagenaria Breviflora leaves on the mild steel surface was
estimated by fitting inhibitor concentration-averaged surface coverage data to the linearised Langmuir isotherm
model. Mean θ values were computed for each unique inhibitor concentration class: 31 ppm (θ = 0.2868, Run
19), 100 ppm (θ = 0.4793, mean of Runs 9, 11, 14, 16), 200 ppm (θ = 0.6450, mean of Runs 1, 3, 5, 6, 12, 13),
300 ppm (θ = 0.7203, mean of Runs 2, 4, 8, 18), and 368 ppm (θ = 0.7701, Run 10). Hence, the linear regression
of C/θ versus C yielded:
The slope of 1.0823, close to unity, validates the Langmuir monolayer assumption and confirms negligible lateral
interactions between co-adsorbed Lagenaria Breviflora leaf extract molecules. The y-intercept of 0.0880 g L⁻¹
clearly visible at C = 0 on the linear plot (Figure 3.4a) equals 1/K
ads
, yielding K
ads
= 11.3649 Lg⁻¹. This high
equilibrium adsorption constant reflects the strong affinity of Lagenaria Breviflora leaf extract phytoconstituents
for the mild steel surface, attributable to the multidentate anchoring capacity of cucurbitacins, flavonoids, and
phenolic acids bearing multiple electron-donor centres. The derived ΔG°
ads
= −16.24 kJmol⁻¹ confirmed a
spontaneous adsorption with a mixed physisorption–chemisorption mechanism, consistent with the activation
energy analysis. The full Langmuir isotherm parameters are presented in Table 4.
Table 3.2 Langmuir adsorption isotherm parameters for the inhibition of mild steel
Adsorption Isotherm
Slope
Intercept (g L⁻¹)
R²
K(Lg⁻¹)
ΔG°ads (kJ mol⁻¹)
Langmuir
1.0823
0.0880
0.9950
11.3649
−16.24
Page 1041
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
Figure 3.4. (a) Linearised Langmuir adsorption isotherm plot (C/θ vs. C (b) Random Forest feature importance
scores
Random Forest Machine Learning Feature Importance Analysis
Random Forest ensemble regression was applied to the complete 20-run CCD dataset using the exact
experimental IE and computed Rc values. These were used to provide a non-parametric, model-agnostic ranking
of the three process variables with respect to their influence on corrosion responses. The RF model achieved
training R² values of 0.9211 Inhibitor Efficiency (IE) and 0.9160 Corrosion rate (Rc), confirming adequate model
fidelity for the compact, 20-observation dataset. Feature importance results were presented in Table 3.3 and
Figure 3.4b.
Inhibitor concentration (Factor A) dominated the variance landscape for both response variables, accounting for
77.54% of explained variance in IE and 74.20% in Rc under the Gini-based MDI criterion. These rankings were
confirmed by permutation importance analysis (100 repetitions) and Gradient Boosting Regression (GBR
importances: IE = 93.00%, Rc = 85.10%), with all three methods consistently identifying inhibitor concentration
as the primary driver. The mechanistic basis for this dominance is straightforward: surface coverage θ rises
sharply from 0.2868 at 31 ppm to 0.7701 at 368 ppm following Langmuir isotherm kinetics, generating a steep,
nonlinear inhibitor concentration-response gradient that accounts for the majority of experimental variance.
However, immersion time (Factor B) ranked as the second most influential variable for IE (12.01%) and Rc
(11.80%), reflecting the progressive mass loss accumulation with time observed in the kinetic study. Likewise,
Temperature (Factor C) contributed 10.45% of IE variance and 14.00% of Rc variance; its slightly greater
influence on Rc relative to IE is mechanistically consistent with Arrhenius acceleration of the blank corrosion
rate, which amplifies the absolute Rc response while the inhibitor film itself remains thermally stable.
Meanwhile, these results of model terms’ significance are consistent with ANOVA results obtained earlier.
Notably, the RF analysis revealed that all three variables made statistically meaningful contributions to the
response variance. This result highlights the value of the multi-variable CCD framework over single-variable
Page 1042
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
experiments and confirms that neither time nor temperature effects should be neglected in field inhibitor
deployment protocols.
Table 3.3 Random Forest feature importance scores (%) for inhibitor efficiency (IE) and corrosion rate (Rc)
Process Variable
RF Importance: IE (%)
RF Importance: Rc (%)
Inhibitor Concentration (ppm)
77.54
74.20
Immersion Time (h)
12.01
11.80
Temperature (K)
10.45
14.00
CONCLUSIONS
This study established and characterised the corrosion inhibition performance of the xylene extract of Lagenaria
breviflora (XEL-B) on mild steel in dilute hydrochloric acid (1 M HCl) through a multi-layered analytical
framework integrating Central Composite Design-structured Response Surface Methodology, Langmuir
adsorption isotherm analysis, three-dimensional adsorption surface modelling, thermodynamic activation
parameter determination, and Random Forest ensemble machine learning. It was observed that XEL-B functioned
as an effective, concentration-dependent green corrosion inhibitor for mild steel in 1 M HCl, with inhibitor
efficiency ranging from 28.68% at 31 ppm to a maximum of 77.01% at 368 ppm, as directly measured across
the 20-run CCD experimental matrix. The fitted quadratic CCD-RSM model was highly statistically significant
(F = 55.81, p < 0.0001) with a non-significant lack of fit (p = 0.2730), confirming its predictive validity. The
generated model equation, accurately described inhibitor efficiency as a function of concentration (x), time (y),
and temperature (z). Furthermore, XEL-B adsorption on mild steel obeyed the Langmuir monolayer isotherm (R²
= 0.9950), with K
ads
= 11.3649 Lg⁻¹ and ΔG°
ads
= −16.24 kJ mol⁻¹ confirming spontaneous, thermodynamically
favourable mixed physisorption-chemisorption adsorption. The three-dimensional Langmuir adsorption surface
corroborated endothermic adsorption behaviour with predicted q
sim
reaching ~477 mg g⁻¹ at Ce = 2000 mg L⁻¹
and T = 100 °C. In addition, activation energy analysis confirmed Ea
inh
= 59.4 kJ mol⁻¹ > Ea
blank
= 43.8 kJ mol⁻¹,
establishing that the adsorbed XEL-B film erects a kinetic energy barrier that substantially retards the
electrochemical dissolution of iron in acidic media. Furthermore, Random Forest machine learning
independently corroborated the ANOVA findings, ranking inhibitor concentration as the dominant process
variable (IE: 77.54%; Rc: 74.20%), followed by immersion time (IE: 12.01%; Rc: 11.80%) and temperature (IE:
10.45%; Rc: 14.00%), with Gradient Boosting cross-validation confirming concordant rankings. This provided
a data-driven basis for prioritising inhibitor dosage as the primary lever in field deployment optimisation.
Collectively, these findings position XEL-B as a promising, biodegradable, and cost-effective alternative to
conventional synthetic corrosion inhibitors in dilute acid service environments. Future work will encompass
electrochemical impedance spectroscopy, potentiodynamic polarisation measurements, and density functional
theory (DFT) molecular orbital calculations on XEL-B phytoconstituent adsorption geometries on Fe(110)
surfaces to further elucidate the electronic and molecular mechanisms underlying the observed inhibition
performance.
REFERENCES
1. Adesina, O. S., Ogundipe, O. B., Ajewole, J. B., Sanyaolu, O. O., Durugbol, J., Adekanye, T. A.,
Olabanji, T. S., Alao, O. P., & Dada, T. J. 2025. Corrosion Challenges, Monitoring Techniques, and
Mitigation Strategies in The Oil and Gas Industry: A Critical Review. Journal of Science and Technology
Research,7, pp. 267–278.
2. Alamiery, A. A., Isahak, W. N. R. W., Aljibori, H. S. S., Al-Asadi, H. A., & Kadhum, A. A. H. (2021).
Effect of the structure, immersion time and temperature on the corrosion inhibition of 4-pyrrol-1-yl-N-
(2,5-dimethyl-pyrrol-1-yl)benzoylamine in 1.0 M HCl solution. International Journal of Corrosion and
Scale Inhibition, 10(2), 700–713. https://doi.org/10.17675/2305-6894-2021-10-2-14
Page 1043
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue III, March 2026
3. Alao, A. O., Popoola, A. P., Dada, M. O., & Sanni, O. (2022). Utilization of green inhibitors as a
sustainable corrosion control method for steel in petrochemical industries: A review. Frontiers in Energy
Research, 10, Article 1063315.
https://doi.org/10.3389/fenrg.2022.1063315
4. Al Otaibi, N., & Hammud, H. H. (2021). Corrosion inhibition using Harmal leaf extract as an eco-friendly
corrosion inhibitor. Molecules, 26(22), 7024. https://doi.org/10.3390/molecules26227024
5. Al-Baghdadi, S., Gaaz, T. S., Al-Adili, A., Al-Amiery, A. A., & Takriff, M. S. (2021). Experimental
studies on corrosion inhibition performance of acetylthiophene thiosemicarbazone for mild steel in HCl
complemented with DFT investigation. International Journal of Low-Carbon Technologies, 16(1), 181–
188.
https://doi.org/10.1093/ijlct/ctaa050
6. Bandeira, R. M., Lima, F. P., Nunes, M. S., dos Santos, E. C., dos Santos Júnior, J. R., de Matos, J. M.
E., Feitosa, C. M., Rai, M., Bhattarai, S., & Das Mulmi, D. (2025). The green plant-based corrosion
inhibitors—a sustainable strategy for corrosion protection. Surface Science and Technology, 3, Article
19.
https://doi.org/10.1007/s44251-025-00019-x
7. Barbu, C. A., Fierascu, I., Semenescu, A., & Cotrut, C. M. (2025). Critical Review Regarding the
Application of Plant Extracts as Eco-Friendly Corrosion Inhibitors—A Sustainable Interdisciplinary
Approach. Molecules, 30(18). https://doi.org/10.3390/molecules30183722
8. Cappelli, F., Castronuovo, G., Grimaldi, S., & Telesca, V. (2024). Random Forest and Feature
Importance Measures for Discriminating the Most Influential Environmental Factors in Predicting
Cardiovascular and Respiratory Diseases. International Journal of Environmental Research and Public
Health, 21(7), 867.
https://doi.org/10.3390/ijerph21070867
9. Iroha, N.B., Akaranta, O. Experimental and surface morphological study of corrosion inhibition of N80
carbon steel in HCl stimulated acidizing solution using gum exudate from Terminalia Mentaly, 2020. SN
Applied Sciences, 2(1514).
https://doi.org/10.1007/s42452-020-03296-8
10. Ituen, E., Singh, A., Yuanhua, L., & Akaranta, O. (2021). Biomass-mediated synthesis of silver
nanoparticles composite and application as green corrosion inhibitor in oilfield acidic cleaning fluid.
Cleaner Engineering and Technology, 3, 100119.
https://doi.org/10.1016/j.clet.2021.100119
11. Kania, H., 2023. Corrosion and Anticorrosion of Alloys/Metals: The Important Global Issue. Coatings,
13(2), https://doi.org/10.3390/coatings13020216
12. Kumari, P., & Lavanya, M. (2022). Plant extracts as corrosion inhibitors for aluminum alloy in NaCl
environment - Recent review. Journal of the Chilean Chemical Society, 67(2).
https://doi.org/10.4067/S0717-97072022000205490
13. Mohammad, A. & Jafar, M., 2020. Global Impact of Corrosion: Occurrence, Cost and Mitigation. Global
Journal of Engineering Science. 5(4): https://doi.org/10.33552/GJES.2020.05.000618
14. Yuan, X., Liu, S., Feng, W. & Dauphin, G., 2023, Feature Importance Ranking of Random Forest-Based
End-to-End Learning Algorithm. Remote Sensing, 15, 5203. https://doi.org/10.3390/rs15215203
15. Zakeri, A., Bahmani, E., & Sabour Rouh Aghdam, A. (2022). Plant extracts as sustainable and green
corrosion inhibitors for protection of ferrous metals in corrosive media: A mini review. Corrosion
Communications, 5, 25–38. https://doi.org/10.1016/j.corcom.2022.03.002
