Seasonally frozen soils are exposed to annual freeze-melt periods, which weakens their mechanical qualities. To precisely characterize the diminishment of soil under various situations, a forecasting system for soil static strength (Ss) has been developed using ML technology. In this study, two machine learning (ML) tactics were designed and validated to evaluate the Ss of seasonally frozen soils, namely the adaptive neuro-fuzzy inference system (ANFIS) and support vector regression (SVR). To find hyper-parameters of models as ideally as possible, the dwarf mongoose optimization algorithm (DMOA) was employed (ANFDW, and SVRDW). Input parameters introduced to the frameworks were water content, below freezing, confining pressure, freeze-melt periods, melting time, and compression degree. The coefficient of determination (R2) quantities for the ANFDW network were discovered to be, respectively, 0.9915 and 0.9965 throughout the training and testing stages. For example, the <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="44147_2025_638_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">${R}^{2}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> for the <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="44147_2025_638_Article_IEq2.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="46" /> </InlineMediaObject> <EquationSource Format="TEX">$ANN$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="italic">ANN</mi> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="44147_2025_638_Article_IEq3.gif" Format="GIF" Height="15" Rendition="HTML" Resolution="72" Type="Linedraw" Width="106" /> </InlineMediaObject> <EquationSource Format="TEX">$PCA-ANN$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>P</mi> <mi>C</mi> <mi>A</mi> <mo>-</mo> <mi>A</mi> <mi>N</mi> <mi>N</mi> </mrow> </math></EquationSource> </InlineEquation> (literature) is 0.97, which is smaller than 0.9915 for the <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="44147_2025_638_Article_IEq4.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="66" /> </InlineMediaObject> <EquationSource Format="TEX">${ANF}_{DW}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">ANF</mi> </mrow> <mrow> <mi mathvariant="italic">DW</mi> </mrow> </msub> </math></EquationSource> </InlineEquation> during the training step. Similarly, in the testing levels, the <InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="44147_2025_638_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">${R}^{2}$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> value is 0.96 compared to 0.9965. Regarding root relative square error (RRSE), ANFDW could receive 0.0925 and 0.0629 lower than 0.1267 and 0.1186 related to the SVRDW in the training and testing sections, respectively. Upon comparing the effectiveness of diverse systems in the literature, although their effectiveness was adequate, the model built in this investigation illustrated superior performance.

Usage of the dwarf mongoose optimization-based ANFIS on the static strength of seasonally frozen soils

期刊论文
出版模式：

开放获取
发表日期：

2025年6月5日

查看全文

Bowen Liu,
Junbin Chen,
Xiaoguang Zhang,
Zhenwei Wang

摘要

Seasonally frozen soils are exposed to annual freeze-melt periods, which weakens their mechanical qualities. To precisely characterize the diminishment of soil under various situations, a forecasting system for soil static strength (S_s) has been developed using ML technology. In this study, two machine learning (ML) tactics were designed and validated to evaluate the S_s of seasonally frozen soils, namely the adaptive neuro-fuzzy inference system (ANFIS) and support vector regression (SVR). To find hyper-parameters of models as ideally as possible, the dwarf mongoose optimization algorithm (DMOA) was employed (ANF_DW, and SVR_DW). Input parameters introduced to the frameworks were water content, below freezing, confining pressure, freeze-melt periods, melting time, and compression degree. The coefficient of determination (R²) quantities for the ANF_DW network were discovered to be, respectively, 0.9915 and 0.9965 throughout the training and testing stages. For example, the ${R}^{2}$ $R^{2}$ for the $ANN$ $ANN$ and $PCA-ANN$ $P C A - A N N$ (literature) is 0.97, which is smaller than 0.9915 for the ${ANF}_{DW}$ ${ANF}_{DW}$ during the training step. Similarly, in the testing levels, the ${R}^{2}$ $R^{2}$ value is 0.96 compared to 0.9965. Regarding root relative square error (RRSE), ANF_DW could receive 0.0925 and 0.0629 lower than 0.1267 and 0.1186 related to the SVR_DW in the training and testing sections, respectively. Upon comparing the effectiveness of diverse systems in the literature, although their effectiveness was adequate, the model built in this investigation illustrated superior performance.