Recent developments in chemical reactive flow of CNTs-hybrid nanomaterial with Soret-Dufour effects using Levenberg-Marquardt technique Authors: Munawar Abbas, Mustafa Bayram, Noor Haval Jamal, Mustafa Inc, Gafur Abdulakimov, Muhammad Shafique, Mohamed Abbas El-Naggar Published: 2026 ABSTRACT In this study, heat and mass transmission in chemical reactive flow of hybrid nanomaterial flowing across a Riga plate with Soret-Dufour effects are studied using integrated numerical computing through artificial neural networks that are back propagated with the Levenberg-Marquardt algorithm (ANN-BPLMA). The suggested model for energy-efficient process cooling in porous media with chemical reaction and Soret-Dufour properties employs a Riga-actuated CNTs-water-based hybrid nanofluid. It has promising applications in current thermal systems. It can be used to improve cooling performance in nuclear reactors, heat exchangers, and electronic device cooling applications requiring high thermal conductivity and accurate heat management. In order to explore the hybrid nanofluid the governing equations provide partial differential equations (PDEs) that mathematically explain temperature, velocity and concentration. The ODEs are generated from these PDEs by using the suitable similarity transformations. The Bvp4c method is used to generate the graphical results. Graphs are employed to describe how various parameters affect temperature, velocity and thermal profiles. The temperature and concentration profiles become more pronounced as the Soret and Dufour numbers increase. Introduction Machine Learning Analysis provides a data-driven technique for modeling and understanding complex engineering processes. Machine learning, which uses advanced learning algorithms with synthetic or experimental datasets, allows for rapid prediction, parameter estimation, and optimization, making it a powerful alternative to traditional computing approaches for a variety of scientific and technical challenges. Khan et al. [1] investigated ANNs (artificial neural networks) examination for entropy-optimized flow of NF (nanofluid) using CCFM (CattaneoChristov flux model). Abbas et al. [2] investigated the modelling of oxytactic bacteria in HNF (hybrid nanofluid) with LTNE effects. Alzahrani et al. [3] investigate the NN (neural network) prediction of magnetohydrodynamic natural convection in THNF (ternary hybrid nanofluids) in a curved enclosure. Hayat et al. [4] examined a NN-based study for porous material caused by radiative entropy. The ML (machine learning) study of HNF flow within a stenosis artery with heat production was examined by Aoudia et al. [5]. The ANN-based temperature study of magnetohydrodynamic flow over a sheet via LMA was observed by Imran et al. [6]. Vinutha et al. [7] explored the study of heat and mass transmission in HNF across a surface using ANNs. Raza et al. [8] investigated the IPNN (intelligent predictive neural network) examination of bioconvection flow of HNF using the CCFM and melting processes. The use of ANNs to optimize magnetohydrodynamic flow in a porous cavity holding HNF was inspected by Tanveer et al. [9]. Raghu et al. [10] use a multi-task NN technique to study the mass transmission prediction of NF flow across a sheet. The use of NF in engineering systems was examined by Aoudia et al. [10] via an ANNs analysis of LTNE effects. Abdal et al. [11] examine the ML study for three-dimensional Casson liquid flow with CCFM in magnetohydrodynamics. Madhukesh et al. [12] used an LMB enhanced NN strategy to optimize the Blasius Rayleigh Stokes flow of aggregated nanoliquid across a semi-infinite plate. Using a 3D magnetohydrodynamic study verified by ANNs, Alharbi and Usma [13] investigated the improved heat transmission properties of a THNF under an angled MF (magnetic field). Shaaban et al. [14] investigated the CCFM and machine learning thermal analysis of HNF for industrial heat transmission uses. The model of three-dimensional flow of MHD non-Newtonian NF across a sheet was observed by Shahbaz et al. [15]. Shahbaz et al. [16] explored the 3D thermally laminated MHD non-Newtonian nanofluids across a stretched sheet. The properties of LTNE on electrophoresis in HFE 7100 nanofluid containing Co3O4 nanoparticles were examined by Liaqat et al. [17] using intelligent predictive NN analysis. Nanofluids are complex colloidal suspensions created by scattering nanoparticles (ranging in size from 1 nm to 100 nm) in conventional base fluids like oil and water. Due to their increased thermal properties, these nanofluids are suitable for a wide range of heat transfer, energy systems, and biological applications. The high ratio of surface area to volume of nanoparticles gives nanofluids their distinctive properties by significantly boosting efficiency of heat transmission even at low solutal. The HNF are liquids that contain two or more dissimilar types of NPs (nanoparticles) dispersed throughout base fluids (water). Such fluids are complex in terms of mass and heat transmission. Hybrid nanofluids are utilized in MHD flows, heat exchangers, and vehicle radiators to improve heat, mass transmission, energy efficacy, and flow behavior in a variety of industrial and technical applications. Razzaq et al. [18] investigated the formation of entropy generation in rheological nanomaterials with CCFM. Liaqat et al. [19] scrutinized the ANN surrogate for Ellis HNF with SDE (Soret-Dufour effects) and Joule heating: Enhance heat and mass transmission. The effects of IPNN on the temperature efficacy of HNF flow were inspected by Khan et al. [20], taking into account elements of heat production and absorption. Khan et al. [21] examined the SATS approach for MF Beyond the traditional technique. Abbas et al. [22] observe the ANNs optimization of gyrotactic germs in water HNF containing CNTs using a modified HCM. The computational investigation of Bodewadt flow and mass transport of THNF over a spinning disc is examined by Vinutha et al. [23]. The magnetohydrodynamic flow of HNF over a wedge affected by melting heat transmission was investigated by Hafeez et al. [24]: engineering application. Tejaswini et al. [25] studied the important effects of hybrid nanoparticles and endothermic CR (chemical reactions) caused by a microchannel with Lorentz forces. An intelligent model for AI-driven analysis of CNTs with wastewater treatment applications was observed by Khan et al. [26]. Based on MHD flow over a wedge with notable effects of SDE, Vinutha et al. [27] observed the temperature conductivity of Fe3O4, SWCNT, MWCNT, and H2O. Ben Hamida et al. [28] examined the temperature study of HNF with gyrotactic microbes using ANNs. Khan et al. [29] examined the computer investigation of parameters for Jeffery liquid flow via an uneven magnetized wavy surface with motile microbes. The durability and various solutions of MHD flow in a DF (Darcy–Forchheimer) permeable surface across a revolving disc were investigated by Mandal and Pal [30]. Using a CattaneoChristov technique, Ahmed et al. [31] observed the features of LF (Lorentz force) and finite-speed heat transmission in buoyant convective flow. Liaqat et al. [32] investigated the BR (Bayesian regularization) optimization method for HNF using the CCFM using wax and sand nanoparticles. The effects of SLIP on mass transmission via CuO-water NF in a channel heated at the center were scrutinized by Srinivas et al. [33]. Anwar et al. [34] analyze the heat transfer aspects in HNF with heat radiations and variable viscosity. Lutera et al. [35] studied the MHD flow of nanofluid with magnetic, and chemical reactions past a moving surface. Heat transfer and HNF flow in a porous cylinder were examined by Malik et al. [36]: MHD influences are examined using a ANNs. Lutera et al. [37] investigated the magnetic field, permeability, and chemical reactions of a dissipative flow impacts of water and silver (Ag) nanoparticles across a porous media. The properties of entropy generation on HNF flow in a staggered cavity were detected by Turabi et al. [38]. Goud et al. [39] analyze the flow of NiZnFe2O4 (Nickel–Zinc Ferrite)—ethylene glycol magneto NF over a permeable sheet with CR. The hybrid nanofluid flow via changing porous region among spinning discs was investigated by Ali et al. [40]. The impact of viscosity dissipation on heat transmission glycol NF flow over a sheet with heat radiation was examined by Dharmendar et al. [41]. The impacts of NPs among two closely flat disk were investigated by Naeem et al. [42] using a thermodynamic analysis of hybrid nanofluid. The computational method of Fe3O4-ethylene glycol mass transmission NF flow with a permeable surface with heat radiation was examined by Reddy et al. [43]. When a temperature difference causes molecular species in a mixture to segregate due to different thermal diffusion coefficients, this is recognized as the Soret consequence. The Soret effect is related to the Dufour effect, which explains how distinct molecular species can be differentiated within a mixture based on their varied thermal diffusion coefficients. Examining the complexities of heat transmission in fluid flow necessitates an accepting of the Soret-Dufour phenomena, which includes both the SDE (Soret-Dufour effects). The fact that these phenomena may be observed in gases, liquids and polymer solutions emphasizes how ubiquitous they are. Jayaprakash et al. [44] explored the temperature radiation, SDE on HNF flow under permeable surface. Khan et al. [45] investigated the Hall current–modified MHD peristaltic mass and heat transmission of Jeffrey liquid. Farooq et al. [46] scrutinized the non-similar investigation of bioconvection NF on a sheet under the influence of SDE. The computational research of SDE in bioconvective flow via a converging-diverging channel was explored by Khan et al. [47]. Islam et al. [48] inspected the lie scaling changes in order to analyses magnetohydrodynamic flow with radiation, over a convective surface during triple diffusion. Sudarmozhi et al. [49] analyze the heat production in dual convection Non-Newtonian magnetohydrodynamic Darcy's flow with SDE. The impacts of viscosity dissipation, radiation, and slip on 3D spinning disc was analyzing by Sivanandam and Alqurashi [50]. Heat transport by heat radiation has a considerable impact on the formation of high temperatures. The many technological processes operate at high thermal, and a thorough kind of radiative heat transmission is crucial for manipulating the equipment. Radiation transfer is analogous to convective heat transfer in various applications, depending on the geometry and surface qualities. There are a few complications while studying radiative heat transfer fluid flow. To begin, when radiative heat transfer occurs, radiation diffuses in the periphery of a system as well as at system boundaries. As a result, prophesy of fluid amalgamation is a difficult task. Second, the heat radiation term in the energy equation renders the equation nonlinear. Third, the coefficient of amalgamation of attractive fluids is strongly dependent on wavelength. Due to such complications, the role of energy on convecting fluid flow has been investigated using logically simplified models. The effects of radiation on Magneto Casson liquid flow on a sheet in a DF with viscous dissipation was analyze by Mangamma et al. [51]. The basic study of MHD flow of HNF in porous surface: Applications of heat radiation and NPs was observed by Ali et al. [52]. The effects of heat transmission NF flow over a flat surface were investigated by Reddy et al. [53]. Khan et al. [54] examined how heat and mass transport in MHD Maxwell hybrid nanofluids flow over porous sheets were influenced by repeated slips. Goud et al. [55] inspected the dynamics of Casson liquid flow over a porous wedge under the outcome of temperature field by numerical analysis. A computational work by Jayaprakash et al. [56] examined the properties of heat radiation, SDE on magnetohydrodynamic mixed convective Maxwell HNF flow under permeable media. The increased solutal and heat fluxes using Reiner-Rivlin material convective flow was analyze by Khan et al. [57]. In their investigation of the transformative energy flow, Sudarmozhi et al. [58] MHD's effect occurs when free convective heat transport with radiation is present. Choudhary et al. [59] investigated the mass-based HNF approach for temperature field study of MHD flow over a wedge embedded in a permeable media. Convective magnetohydrodynamic flow of Casson blood THNF over a spinning rigid disk was inspected by Gahgah et al. [60] in relation to form factor and temperature radiation. These studies examine the Soret-Dufour impacts affect mass transfer in the flow of a HNF containing SWCNTs and MWCNTs spread in water across a Riga plate. The current study is intended to fill this gap. The main goal of this research is to recognize the elements that improve the mass transmission of base fluids (water) suspended in SWCNTs and MWCNTs. The PDEs are resolved mathematically after transforming in the form of ODEs. By applying the shooting method in Mathematica software graphs have been plotted and highlighted in briefly to demonstrate how several essential physical factors influenced the HNF rate of heat transfer. The suggested model for energy-efficient process cooling in permeable media with chemical reaction and SDE employs a Riga-actuated CNTs-water-based hybrid nanofluid. It has promising applications in current thermal systems. It can be used to improve cooling performance in heat exchangers, nuclear reactors, and electronic device cooling applications requiring high thermal conductivity and accurate heat management. Furthermore, the incorporation of porous media and chemical reactions makes it applicable to fuel cells, and energy storage devices, catalytic reactors, while the SDE improve mass transmission processes in biomedical, and environmental engineering uses. Mathematical formulation The impact of the thermal radiation on 2D steady and incompressible flow of MWCNTs+SWCNTs/Water base HNF over Rig plate with Soret-Dufour impacts. The flow consists of SWCNTs and MWCNTs distributed in water across a Riga plate embedded with a porous material. To maximize heat and mass transmission, space is considered changeable. Fig. 1 displays a physical representation of the proposed model.Fig. 1Physical flow description.Fig 1 dummy alt text Physical flow description. The governing equation are as given [61,62].(1)∂u∂x+∂v∂y,(2)(u∂u∂x+v∂u∂y)ρhnf=μhnf∂2u∂y2+πJ0M*8ρhnfe−(π/b)y−μhnfK1u+(gβ)hnf[βt(−T∞+T)−βc(−C+C∞)],(3)(v∂T∂y+u∂T∂x)(ρcp)hnf=khnf∂2u∂y2+16σ*T∞33k*∂2T∂y2+ρDmKtCs(ρCp)hnf∂2C∂y2,(4)u∂C∂x+v∂C∂y=Dm∂2C∂y2−+DmKtTm∂2T∂y2. Boundary conditions [61,62].(5)μhnf∂u∂y|y=0=∂σ∂x|y=0=−σ0(γC∂C∂x|y=0+γT∂T∂x|y=0)aty=0,(6)C=Cox2+C∞,T=Tox2+T∞,v=0,aty=0,(7)C→C∞,T→T∞,u→0,asy→∞. The ST is assumed to be linearly related to thermal and solutal variables.(8)σ=−γT(T−T∞)+σ0−γC(C−C∞0),(9)γC=−(∂σ∂C)C=c∞andγT=−(∂σ∂T)T=T∞. In this work (u,v) represents components of velocity in (x,y) direction, gβ[βt(T−T∞)+βc(C−C∞)] represents buoyancy force, K1 represents permeability of porous medium, ρDmKtCs(ρCp)hnf∂2C∂y2 represents Dufour effect and DmKtTm∂2T∂y2 represents Soret effect. The similarity transformations are given [36].(10)η=ycνf,v=−cνfF(η),u=cxF′(η),(11)C=C∞+Cox2ϕ(η),T=T∞+Tox2θ(η), Eq. (1)-Eq. (7) are transformed into ODEs by applying Eq. (8).(12)D1F″′+D2(FF″−Fˊ2)+Hae−βη+D3Mc(θ+Nbϕ)−D1λF′=0,(13)(D4+Rd)θ″+W5Pr(Fθ′−2F′θ)+Duϕˊˊ=0,(14)ϕˊˊ+(ϕ′F−2F′ϕ)−RcScϕ+ScSrθ″=0, The BCs are:(15)D1F″=−(1+Ma)2Mn,θ=1,F=0,ϕ=1,η=0,(16)F′→0,θ→0,ϕ→0,η→∞. Here Q=Q0(ρCp)fc Heat source parameter, λ=νfK1c porosity parameter, Nb=βc(Cw−C∞)βt(Tw−T∞) buoyancy concentration parameter, Du=ρDmKt(Cw−C∞)Cs(ρCp)hnfνf(Tw−T∞) Dufour effect, Sc=Dmνf Schmidt parameter, Ha=πJ0M*8ρfUwc Hartman number,Pr=νfαf Prandtl parameter, Mc=gββt(Tw−T∞)Uwc Mixed convection parameter, Sr=DmKt(Tw−T∞)Tmνf(Cw−C∞) Soret effect. Table 1, Table 2 display the basic aspects of HNF.Table 1The experimental values of SWCNTs, MWCNTs and H2O [[63], [64], [65]].Table 1 dummy alt textMaterialρkCpσH2O997.10.61341795.5×10−6SWCNTs26003.74251.1×107MWCNTs1600300079659.6×106Table 2Thermophysical properties for hybrid nanofluid [65].Table 2 dummy alt textProperty nameSymbolRelationAbsolute viscosityD1=μhnfμf(1−φ1)−2.5(1−φ2)−2.5.DensityD2=ρhnfρf(1−φ2)[(−φ1+1)+φ1(ρs1ρf)]+(ρs2ρf)φ2.Heat capacityD4=(ρCp)hnf(ρCp)f(−φ2+1)[(−φ1+1)+((ρCp)s1(ρCp)f)ϕ1]+((ρCp)s2(ρCp)f)φ2.Thermal conductivityD3=khnfkfkhnfknf=(1−φ2)+2φ2(ks2ks2−knf)In(ks2+knfknf)(1−φ2)+2φ2(knfks2−knf)In(ks2+knfknf),knfkf=(1−φ1)+2φ1(ks1ks1−kf)In(ks1+kfkf)(1−φ1)+2φ1(kfks1−kf)In(ks1+kfkf). Table 1The experimental values of SWCNTs, MWCNTs and H2O [[63], [64], [65]].Table 1 dummy alt textMaterialρkCpσH2O997.10.61341795.5×10−6SWCNTs26003.74251.1×107MWCNTs1600300079659.6×106 The experimental values of SWCNTs, MWCNTs and H2O [[63], [64], [65]]. Table 2Thermophysical properties for hybrid nanofluid [65].Table 2 dummy alt textProperty nameSymbolRelationAbsolute viscosityD1=μhnfμf(1−φ1)−2.5(1−φ2)−2.5.DensityD2=ρhnfρf(1−φ2)[(−φ1+1)+φ1(ρs1ρf)]+(ρs2ρf)φ2.Heat capacityD4=(ρCp)hnf(ρCp)f(−φ2+1)[(−φ1+1)+((ρCp)s1(ρCp)f)ϕ1]+((ρCp)s2(ρCp)f)φ2.Thermal conductivityD3=khnfkfkhnfknf=(1−φ2)+2φ2(ks2ks2−knf)In(ks2+knfknf)(1−φ2)+2φ2(knfks2−knf)In(ks2+knfknf),knfkf=(1−φ1)+2φ1(ks1ks1−kf)In(ks1+kfkf)(1−φ1)+2φ1(kfks1−kf)In(ks1+kfkf). Thermophysical properties for hybrid nanofluid [65]. The physical variables related to drag force, mass and heat transmission behaviors are Nusselt number Nu, Sherwood number Sh, and skin-friction constant Cf. These variables are defined as [61,62].(17)Cf=τwρfUw2,Nu=xqwkf(Tw−T∞),Sh=xqmDm(Cw−C∞),(18)τw=μhnf∂u∂y,qw=−(kthnf+16σ*T∞33k*)∂T∂y|y=0,qm=−Dm∂C∂y,Rex=xUwνf,(19)CfRex0.5=D1F″(0),NuRex−0.5=−(D4+Rd)θ′(0),ShRex−0.5=−ϕ′(0). Methodology The BVP4C method is very helpful since it solves nonlinear BVPs in fluid mechanics and heat transmission in an accurate and efficient manner. It uses a comparison approach with adaptive mesh refinement to increase correctness around steep gradients or singularities. Furthermore, BVP4C is resilient, simple to implement in MATLAB, and capable of handling complex boundary conditions, making it an effective tool for studying real-world engineering and scientific challenges. Artificial neural networks In order to solve difficult real-world problems, MLAs have been established in a variety of academic disciplines, such as engineering, science, and mathematics. By completing tasks or examples, the Supervised Learning Algorithm (SLA) simulates how humans learn. To aid in this process, a training dataset—also known as a teaching signal—is provided to the neural network (NN). In order to improve the NN's effectiveness in learning the desired mapping, the model can modify its internal weights by back-propagating the error between the expected and actual outputs during training. This section provides a quick summary of the findings for the suggested Artificial Neural Network with Levenberg-Marquardt technique. The reliable and smooth convergence of the training curves across all epochs provides additional evidence of the artificial neural networks that are back propagated with the Levenberg-Marquardt algorithm (ANN-BPLMA) framework's efficacy and operational efficiency. This behavior shows that a stable equilibrium point has been reached by the algorithm. Because it represents the model's dependable and consistent learning behavior during the training phase, this stability is essential. The progressive decrease in MSE values shown in Table 3 and the associated visual results demonstrate that the model consistently produces better results, as confirmed by the performance assessment. The idea that lower error values correlate to increased accuracy and greater predictive capabilities of the ANN-BPLMA configuration is amply supported by the negative association between performance and MSE. This finding supports the hypothesis that more model validity is indicated by a lower mean squared error. The stability and optimal performance attained during training are further highlighted by the continuous and well-behaved convergence lines. Table 3 illustrates that higher precision and sharper prediction quality are indicated by decreased MSE values. An enlarged perspective of the pattern of deviations around the zero-baseline following training is provided by the error histogram analysis shown in Figs. 2b, 3b and 4b. The degree of discrepancy between the actual and expected outputs is made clearer by this diagnostic image. The BPLM technique shows that there is a significant association among the real reference values and the anticipated outcomes in all three case studies, with the central tendency of the error bins shifting slightly away from zero. The suggested fluid dynamic model is thoroughly evaluated by the fitness evaluation plots displayed in Figs. 2c, 3c and 4c. The differences between calculated solutions and benchmark reference data are examined in these graphics. The model's capacity to consistently trace the reference behavior is demonstrated by the smooth overlap between the target profiles and those produced by the BPLM solver in every case. The neural network architecture's structural soundness and predictive power are validated by the strong correlation between these curves. The great degree of agreement confirms that the BPLM model can faithfully represent and reproduce the complex features of the fluidic system being studied. The alignment seen in the fitness plots demonstrates the established method's resilience and validates its reliability in simulating the configurations under investigation. The accuracy, stability, and convergence properties of ANN-BPLMA are further confirmed in Figs. 2d, 3d and 4d. As the network gets closer to ideal performance, these figures illustrate how the Mu and gradient values change. The gradient gradually decreases as learning advances, eventually attaining an insignificant value that, when it drops below the designated cutoff level, indicates that training is finished. The adaptive Mu factor modifies itself according to the convergence rate in response to the shifting error landscape. To complete the validation process, RA (regression analysis) is performed to show the connection among actual target values and predicted outputs. Every data point should ideally be on the diagonal reference line in perfect alignment. Regression plots for training, validation, and testing data separately are displayed in Figs. 2e, 3e and 4e, demonstrating that the network maintains robust predictive function across all data partitions.Table 3An evaluation of the artificial neural network's numerical performance.Table 3 dummy alt textScenariosEpochsPerformanceTrainingTestingValidationI1572. 79 × E−073.25 × E−071.26 × E−071.51 × E−07II9233. 74 × E−103.84 × E−103.81 × E−103.16 × E−10III2651.20 × E−071.87 × E−082.30 × E−075.51 × E−07Fig. 2An illustration of how the ANN-BPLMA model performed in scenario I.Fig 2 dummy alt textFig. 3The performance of ANN-BPLMA in scenario II is shown graphically.Fig 3 dummy alt textFig. 4An illustration of how the ANN-BPLMA forecasts in scenario III.Fig 4 dummy alt text Table 3An evaluation of the artificial neural network's numerical performance.Table 3 dummy alt textScenariosEpochsPerformanceTrainingTestingValidationI1572. 79 × E−073.25 × E−071.26 × E−071.51 × E−07II9233. 74 × E−103.84 × E−103.81 × E−103.16 × E−10III2651.20 × E−071.87 × E−082.30 × E−075.51 × E−07 An evaluation of the artificial neural network's numerical performance. An illustration of how the ANN-BPLMA model performed in scenario I. The performance of ANN-BPLMA in scenario II is shown graphically. An illustration of how the ANN-BPLMA forecasts in scenario III. Results and discussions The proposed model for energy-efficient process cooling in permeable medium with chemical reaction and SDE (Soret Dufour effects) uses a Riga-actuated CNTs-water-based hybrid nanofluid. It has promising applications in modern thermal systems. It can be utilized to enhance cooling performance in heat exchangers, nuclear reactors, and electronic device cooling applications that require high thermal conductivity and precise heat control. Furthermore, the incorporation of porous media and chemical reactions makes it suitable for catalytic reactors, fuel cells, and energy storage devices, whereas the Soret and Dufour effects increase heat transfer processes in biomedical, chemical, and environmental engineering applications. Fig. 5, Fig. 6 illustrates how Ma effects on θ and F′. The HNF velocity profiles are improved by rising Mavalues, as the graphs show. These numbers show a significant drop in the thermal profile as Ma rises. The attraction among surface molecules grows as the bulk magnetization of the fluid yields ST (surface tension) for the particles in the surface layer. Fig. 7. investigates that drag forces and resistance are produced by the porosity parameter λ in the porous media which lowers the fluid F′(η). The F′(η) decreases as drag and resistance rise as the porosity parameter is boosted. Fig. 8. illustrates that how porosity parameter effects on temperature profile θ(η). By raising the porosity parameter λ causes the fluid resistance in the permeable media to decrease, increasing the temperature and storing more heat as a result θ(η) raises. The influence of Φon F′, and θfields of HNF are portrayed in Fig. 9, Fig. 10. When volume fraction parameter’s function, BL width and the thermal field increase while the momentum BL (boundary layer) drop. Additionally, as Φ rises, heat generation rises as well, improving the thermal field and the width of the related BL. This rise in the Φ is coupled with a concurrent drop in the flow profiles of HNF. Fig. 11. describe that when Rd increases it supplies more and more temperature energy to liquid which results increase in fluid’s temperature θ(η). Hence θ(η) increases as Rd raises. The outcome of Rc on ϕis shown in Fig. 12. The ϕ (η) declines as Rc rises. Since of the chemical process that drops the solutal field. The impact of Du on θ,Sr and Nu, on ϕ,Sh,are demonstration in Fig. 13, Fig. 14, Fig. 15, Fig. 16. Fig. 13. demonstration that how does θupsurge as Dufour number Du increases. As Du the rises concentration gradients result in greater heat transfer which causes increase in the θ(η) overall. Fig. 14. show that by increasing the concentration gradient through thermal diffusion the Soret number Sr improves mass transfer. Thus, a higher Sr also results in a higher concentration profile ϕ(η). The heat transmission rate surge as augment the values of Du and Sr. The result of Φon Nu and Sh are showed in Fig. 17, Fig. 18. The heat transmission rate rise as increase the values of Φ while mass transfer decline. The significance of Rd on Rc and Nu on Sh are display in Fig. 19, Fig. 20. The heat transmission rate upsurge as augment the radiation factor. The mass transmission rate decrease as enhances the values of Rc.Fig. 5Result of Maon f′.Fig 5 dummy alt textFig. 6Effects of Maon θ.Fig 6 dummy alt textFig. 7Effect of λ on F′.Fig 7 dummy alt textFig. 8Outcome of λ on θ.Fig 8 dummy alt textFig. 9Influence of Φ on F′.Fig 9 dummy alt textFig. 10Impact of Φ on θ.Fig 10 dummy alt textFig. 11Result of Rd on θ.Fig 11 dummy alt textFig. 12Encouragement of Rc on ϕ.Fig 12 dummy alt textFig. 13Outcome of Du on θ.Fig 13 dummy alt textFig. 14Encouragement of Sr on ϕ.Fig 14 dummy alt textFig. 15Impact of Du on Nu.Fig 15 dummy alt textFig. 16Result of Sr on Sh.Fig 16 dummy alt textFig. 17Influence of Φ on Nu.Fig 17 dummy alt textFig. 18Outcome of Φ on Sh.Fig 18 dummy alt textFig. 19Outcome of Rd on Nu.Fig 19 dummy alt textFig. 20Outcome of Rc on Nu.Fig 20 dummy alt text Result of Maon f′. Effects of Maon θ. Effect of λ on F′. Outcome of λ on θ. Influence of Φ on F′. Impact of Φ on θ. Result of Rd on θ. Encouragement of Rc on ϕ. Outcome of Du on θ. Encouragement of Sr on ϕ. Impact of Du on Nu. Result of Sr on Sh. Influence of Φ on Nu. Outcome of Φ on Sh. Outcome of Rd on Nu. Outcome of Rc on Nu. Conclusions In this research analysis the impact of thermal radiation on flow of HNF across a Riga plate with Soret and Dufour effects. Water is mixed with two dissimilar kinds of carbon nanotubes SWCNTs and MWCNTs. The proposed model for energy-efficient process cooling in permeable medium with Soret-Dufour significance makes use of a Riga-activated CNTs-water-based hybrid nanofluid. It has potential applications in modern thermal systems. It can be utilized to boost cooling performance in heat exchangers, nuclear reactors, and electronic device cooling applications that require high thermal conductivity and precise heat control. Furthermore, the use of porous media and chemical reactions makes it suitable for catalytic reactors, fuel cells, and energy storage devices, while the Soret and Dufour effects increase mass transfer processes in chemical, biomedical, and environmental engineering uses. Followings are the main findings of problem.•The reliability and efficacy of the suggested computational solution are confirmed by the numerical results, which exhibit strong agreement with the results produced by the artificial neural networks that are back propagated with the Levenberg-Marquardt algorithm (ANN-BPLMA).•Fluid velocity increases while thermal profile decline by increasing the Marangoni convection parameter.•Temperature profile and rate of heat transmission upsurges while flow field and rate of mass transmission decrease by growing the values of nanoparticles volume friction.•Fluid temperature upsurges by increasing the Dufour number, thermal radiation parameter and porosity factor.•Concentration profile raises by increasing the Soret number and declines by raising the values of chemical reaction factor.•Nusselt number surges for the maximum values of Dufour number and radiation parameter.•Sherwood number upsurges as the values of Soret number rises but decreases as the values of Sc and τ enhances.•The model's improved capacity to produce precise and trustworthy predictions is reflected in a lower MSE assessment. •The reliability and efficacy of the suggested computational solution are confirmed by the numerical results, which exhibit strong agreement with the results produced by the artificial neural networks that are back propagated with the Levenberg-Marquardt algorithm (ANN-BPLMA). The reliability and efficacy of the suggested computational solution are confirmed by the numerical results, which exhibit strong agreement with the results produced by the artificial neural networks that are back propagated with the Levenberg-Marquardt algorithm (ANN-BPLMA). •Fluid velocity increases while thermal profile decline by increasing the Marangoni convection parameter. Fluid velocity increases while thermal profile decline by increasing the Marangoni convection parameter. •Temperature profile and rate of heat transmission upsurges while flow field and rate of mass transmission decrease by growing the values of nanoparticles volume friction. Temperature profile and rate of heat transmission upsurges while flow field and rate of mass transmission decrease by growing the values of nanoparticles volume friction. •Fluid temperature upsurges by increasing the Dufour number, thermal radiation parameter and porosity factor. Fluid temperature upsurges by increasing the Dufour number, thermal radiation parameter and porosity factor. •Concentration profile raises by increasing the Soret number and declines by raising the values of chemical reaction factor. Concentration profile raises by increasing the Soret number and declines by raising the values of chemical reaction factor. •Nusselt number surges for the maximum values of Dufour number and radiation parameter. Nusselt number surges for the maximum values of Dufour number and radiation parameter. •Sherwood number upsurges as the values of Soret number rises but decreases as the values of Sc and τ enhances. Sherwood number upsurges as the values of Soret number rises but decreases as the values of Sc and τ enhances. •The model's improved capacity to produce precise and trustworthy predictions is reflected in a lower MSE assessment. The model's improved capacity to produce precise and trustworthy predictions is reflected in a lower MSE assessment. Abbreviations CsConcentration of solubilityu,vVelocity’s componentsM*Magnetic fieldHaHartman numberDmDiffusivity coefficientgβAcceleration due to gravityK1Permeability of porous mediaβcSolutable expansion coefficientsλPorosity parameterβtThermal expansion’s coefficientMcMixed convection parameterCpSpecific heatPrPrandtl parameterTFluid temperatureNbBouyancy concentration parameterSrSoret numberDuDufour numberScSchmidt numberTwWall temperatureT∞Ambient temperatureCwWall concentrationC∞Ambient concentration CsConcentration of solubilityu,vVelocity’s componentsM*Magnetic fieldHaHartman numberDmDiffusivity coefficientgβAcceleration due to gravityK1Permeability of porous mediaβcSolutable expansion coefficientsλPorosity parameterβtThermal expansion’s coefficientMcMixed convection parameterCpSpecific heatPrPrandtl parameterTFluid temperatureNbBouyancy concentration parameterSrSoret numberDuDufour numberScSchmidt numberTwWall temperatureT∞Ambient temperatureCwWall concentrationC∞Ambient concentration CRediT authorship contribution statement Munawar Abbas and Mustafa Bayram: Supervision, Methodology, Conceptualization. Noor Haval Jamal: Conceptualization; Mustafa Inc and Gafur Abdulakimov: Methodology, Formal analysis., Muhammad Shafique: Methodology, Conceptualization. Mohamed Abbas El-Naggar: Writing – review & editing. Declaration of competing interest The authors have no relevant financial or non-financial interests to disclose.