Bayesian regularization technique for chemical reactive flow dynamics of hybrid nanomaterial using Xue model: Chemical reactor optimization Authors: Waleed Mohammed Abdelfattah, Munawar Abbas, Mustafa Bayram, Farkhod Rakhmonov, Humaira Kanwal, Ali Hasan Ali Published: 2026 ABSTRACT This study inspects the flow dynamics of Stefan blowing effects and heat radiation on chemical reactive flow of hybrid nanofluid using a Riga plate in the existence of activation energy. To model and address the intricate nonlinear issue, artificial neural networks (ANNs) that have been trained using the Bayesian regularization back-propagation strategy (ANN-BRS) are utilized. Gyrotactic microbes are integrated with nanoparticles in various applications, including microfluidic platforms, bio-microsystems on chip-scale devices, enzyme-based biosensors, bacteria-driven micromixers, microbial fuel cells, and other micro-engineered systems, to enhance thermal efficacy. This method can also help environmental engineering by improving wastewater treatment procedures by allowing microbes to more effectively degrade pollutants. It raises the production of biofuel in the realm of renewable energy by advancing the creation of more effective photo bioreactors. Material scientists to produce regulated nanostructured materials with consistent compositions and thermal characteristics may use this idea. A higher Stefan blowing factor leads to a higher flow profile. The concentration field decreases as the values of the chemical reaction parameter increases. Nomenclature Nusselt Number Streams function Average temperature of the Fluid Dynamic viscosity Pecklet number Dimension less constant Electrical conductivity Biconvection Schmidt number Thermal conductivity Ambient concentration Cartesian coordinates Fluid density Thermal diffusivity Chemical reaction Stefan blowing parameter Hartman number Ambient Concentration Gyrotactic microorganisms Sherwood Numb Diffusion Coefficient Chemical reaction parameter Width of electrodes and magnets Activation energy parameter Microorganism concentration Lorentz force Volume friction of nanoparticles Magnetic strength Prandtl number Local density of bioconvection Liquid temperature Kinematic viscosity Heat generation parameter Ambient temperature Schmidt number The intensity of the current in electrodes Velocity fields Nanoparticles Darcy-Forchheimer flow Machine learning Local thermal non-equilibrium conditions Nanofluid Tri-hybrid nanofluids -Hybrid nanofluid Cattaneo-Christov heat flux model Boger nanofluid Introduction For complex fluid flow applications, ML (machine learning) approaches such as neural network and regression models deliver pressure, and heat predictions, real-time flow, providing a high-processing-power substitute for traditional CFDs (computational fluid dynamics) modeling. Thus, without compromising accuracy, ML may enhance aerodynamic structures, and heat exchangers. Additionally, machine learning-based models can enhance turbulence forecasts and surpass traditional turbulence models in correctness by learning through realistic modeling, such as Direct Numerical Simulation. By finding critical elements that optimize efficacy and decrease energy losses, ML advances the study of fluid flow in uses such as industrial cooling systems and aerodynamic optimization. Abdal et al. [1] scrutinize the ML examination of CCHFM (Cattaneo-Christov heat flux model) in Casson fluid flow in three-dimensional magnetohydrodynamics by the linear regression method. Shaaban et al. [2] examine enzyme-based biosensor uses for the Boger nanomaterial flow as gyrotactic microbes were present using ML based on the BR (Bayesian regularized) approach. Adogbeji et al. [3] inspect thermal uses of tri-hybrid nanofluids: machine learning prediction, magneto-hydrodynamics, and stability. Abbas et al. [4] examine LTNECs in a THNF (trihybrid nanofluid's) flow and heat radiation is solved using ANNs (artificial neural networks) enhanced by the BR method. Shah et al. [5] examined the application of Bayesian distributed NN (neural networks) for machine learning study of the behavior of gyrotactic microbes in a hydromagnetic bio-convected nanofluid. Hussein et al. [6] investigate the development of ANNs the Boger HNF (hybrid nanofluid) Darcy-Forchheimer flow is described by the Cattaneo-Christov theory. The heat rheological response of biological nanofluids with thermophoresis diffusion processes was studied by Abdelfattah et al. [7] using semi-analytical and intelligent computer paradigms. The numerical optimization of the Al2O3-TiO2/Water HNF thermal performance was observed by Arif et al. [8]. The slip impacts of solarized BNF with activation energy were examined by Liaqat et al. [9] using a Bayesian regularized ANNs technique based on the LMA (Levenberg-Marquardt algorithm). Qayyum et al. [10] investigated a viable approach for higher-order BVPs (boundary value problems). Abbas et al. [11] investigated the solutal and temperature study of oxytactic microorganisms in slip flow of THNF with AE (activation energy) using an ANNs. Vinutha et al. [12] investigate the application of NN to optimize heat transfer HNF (hybrid nanofluids) over a surface. Liaqat et al. [13] conduct an intelligent predictive NN examination of LTNE's effects on thermophoresis in nanofluid. Arif et al. [14] scrutinized the use of artificial intelligence to simulate mass transmission in magnetic HNF as an industrial use. A mixture of several nanoparticles floating in a base fluid is called a HNF (hybrid nanofluid). Both metallic and nonmetallic nanoparticles are available. The hybrid component of the nanofluid may have better qualities than individual-component nanofluids due to the synergistic effects produced by the interaction of many nanoparticle kinds. Moyes and Anala [15] examined MHD- driven hybrid nanofluid flow of the third grade over a sheet affected by a heat production. A statistical approach to the HNF flow via a porous convective moving surface with the generation of heat was studied by Wahid et al. [16]. Kaswan et al. [17] explored the flow of an HNF (hybrid nanofluid) with nonlinear heat radiation between two concentric cylinders while minimizing entropy. Rahman et al. [18] scrutinized the thermo mechanics of a radiated hybrid nanofluid interacting with a heating source and MHD. Aquib et al. [19] evaluated the understanding of heat flux in HNF flow across a cylinder. Hadidi et al. [20] scrutinized the BRNN (Bayesian regularized neural network) research of mass transmission in a CMC and water HNF under LTNE. Tejaswini et al. [21] investigated the optimization of HNF flows between moving flat plates and stationary using chemical processes. Liaqat et al. [22] study the BR optimization strategy for HNF with CCHFM utilizing sand NPs. Ali et al. [23] examined the temperature radiation of DFF (Darcy-Forchheimer flow) with CCHFM across a Riga plate using a numerical examination. Ben Hamida et al. [24] used an ANNs to scrutinize the temperature examination of the flow of a HNF having gyrotactic microbes. Vinutha et al. [25] investigated the computational inspection of THNF mass transportation and Bodewadt flow over a spinning disk controlled by waste discharge system. Abbas et al. [26] investigate the ANNs investigation for oxytactic microorganisms in a HNF containing chemical reactions. Ullah et al. [27] study the THNF flow with porosity, MHD, and temperature properties on a 3D stretched surface with CBCs (convective boundary conditions). Heat transmission from a heated entity to its surroundings is described by the dimensionless Stefan parameter, also known as the Stefan number. When working with high-temperature systems or circumstances relating significant thermal variations, the parameter is very important. Researchers may evaluate the effects of heat transmission, identify suitable heat transmission processes, and improve designs for effective temperature management of objects in a variety of uses by comprehending and evaluating the Stefan factor. Haider et al. [28] inspected the unsteady flow of NF across a stretched sheet with a AE (activation energy) and heat radiation is affected by Stefan blowing. Ali et al. [29] assessed the SB (Stefan blowing) influences of heat radiation on flow of bio-convection through a revolving disk in a porous media. Geetha et al. [30] inspect the analysis of mass transmission in an unstable Carreu NF flow in MHD across a surface in a permeable media under Stefan blowing conditions. Reddy et al. [31] scrutinized the flow of a HNF with MHD and Stefan blowing affects heat radiation across a moving thin needle. Joyce et al. [32] examine how double stratification and Stefan blowing have an impact on HNF flow with entropy production and convective barriers. Activation energy is the minimum quantity of energy needed to start a chemical reaction. Mass movement and chemical measures frequently interact in intricate ways. It can be proven utilizing liquid flow, constituent species absorption, and manufacture at various mass transmission speeds. Arrhenius promoted the use of Arrhenius AE in mixture with species-specific chemical reactions for the initial time. The activation energy value has a major impact on the movement of NPs in simple carrier fluids. Wantao et al. [33] evaluated the result of activation energy on mass heat transport in THNF flow over orthogonal permeable discs of different geometries. Goud et al. [34] reviewed the aspects of MHD heat radiation's melting heat transport across a stretched sheet under AE. Fatima et al. [35] scrutinized HNF with AE were analyzed thermally and MHD. Thumma et al. [26] examine the implications of activation energy related to velocity slip in MHD flow over a surface. Norzawary et al. [36] investigated the electrical MHD flow using a modified Buongiorno model via stretched sheet with AE. Padma et al. [37] observed the outcome of AE on MHD flow across flat and parabolic surfaces. Using a modern fractional approach, Khan et al. [38] inspected the effect of a Darcy's medium with heat generation on MHD convective flow. Qayyum et al. [39] use the He-Laplace method to study the transporting waves solutions of extended seventh-order time-fractional KdV models. Saeed et al. [40] investigate and use convection flow in a channel. The main properties of endothermic chemical processes and HNF made via a microchannel with LF (Lorentz forces) were scrutinized by Tejaswini et al. [41]. The impact of CBCs and autocatalytic chemical processes on NF oblique stagnation point flow across a sheet was studied by Vinutha et al. [42]. An unsteady HNF (Cu-UO2/blood) featuring chemical processes and nonlinear temperature radiation across CBCs was studied by Qayyum et al. [43]. Nasir et al. [44] investigated the chemically radioactive couple stress Casson HNF flow on a rotating sphere. Habib et al. [45] use the energy source and Cattaneo-Christov heat flux model to study how chemically reacting NPs improve thermal transmission. Nasir et al. [46] investigated the efficacy of solar radiation on chemical radioactive NF flow across a porous surface with a MF (magnetic field). Habib et al. [47] investigated a machine learning-driven analysis of heat transmission in chemically reactive fluid flow with Soret-Dufour effects. Nasir et al. [48] investigated nanoparticle dynamics in binary CRs (chemical reactions) in magnetized porous media using a computational study. Eswaramoorthi et al. [49] investigated the numerical simulation of CNT nanofluids rotating under thermal radiation and autocatalytic chemical processes. Microbial activity dramatically modifies flow behavior in fluid flow interacting with microorganisms, which is a highly complicated multiphase process. In biomedical settings, pathogens and cellular components alter blood flow; interactions with immune cells and platelets impact vascular function and rheological characteristics. Microbial metabolism has the ability to locally alter thermal properties of natural water bodies, which can affect heat dispersion. For improvements in engineering, environmental, and biological applications, it is crucial to realize these linked fluid-microbial-thermal interactions. Hussain et al. [50] examine examining the basic aspects of nanofluid flow that capturing elements of shear thinning and thickening with microbes' gyrotactic. Ramzan et al. [51] scrutinized MHD flow of a viscoelastic liquid in 3D with mixed convection with gyrotactic microbe physical aspects is determined. Sadighi and Marin [52] explored the in-depth theoretical study of heat-mass transport and biomagnetic fluid flow on a permeable stretching/shrinking cylinder that is host to motile gyrotactic germs. Sankari et al. [53] inspected the analysis of flow dynamics and bioconvective transmission caused by gyrotactic microbes in NFs via a growing surface with solutal and thermal influences. Sandeep et al. [54] investigate the Weissenberg number and the properties of angled MHD repeated slips on the flow of micro-motile organisms. The numerical investigate the consequence of heat source on hybrid liquid flow via a Riga plate is the novel aspect of the current work. The Homotopy analysis method is used in our present investigation to determine the answer. Because of its efficiency, accuracy, flexibility, and ease of use, the shooting technique is a common strategy for resolving Stefan blowing. The result of abundant dimensionless parameters on the concentration, thermal, and flow dispersals have been examined. The specified goal of the research is:➢To regulate the features of Stefan blowing on hybrid nanoliquid.➢Scrutinize the features of the thermal radiation on TBL flow of hybrid nanofluid.➢To determine the result of activation energy on the concentration profile of HNF.➢To assess the importance of gyrotactic microbes in hybrid nanofluid. ➢To regulate the features of Stefan blowing on hybrid nanoliquid. To regulate the features of Stefan blowing on hybrid nanoliquid. ➢Scrutinize the features of the thermal radiation on TBL flow of hybrid nanofluid. Scrutinize the features of the thermal radiation on TBL flow of hybrid nanofluid. ➢To determine the result of activation energy on the concentration profile of HNF. To determine the result of activation energy on the concentration profile of HNF. ➢To assess the importance of gyrotactic microbes in hybrid nanofluid. To assess the importance of gyrotactic microbes in hybrid nanofluid. This model has important implications in bio-thermal and environmental engineering systems, where microorganism dynamics and improved heat transfer are critical. Machine learning analysis of gyrotactic germs in HNF bioconvection flow with AE, based on the Xue model and including Stefan blowing effects, can be used in bioreactors, microbial fuel cells, wastewater treatment, and bio-microsystems. It helps to optimize mass and heat transmission, manage microbe distribution, and increase energy efficiency in bio-inspired cooling technologies, renewable energy systems, and biomedical and environmental processing applications. Mathematical formulation The heat source properties on hybrid nanofluid flow via a Riga plate with AE (activation energy) (Fig. 1). The stretching occurs in the same plane y=0. The layer's movement is presummated to be linear, that means for a stretched surface, uw(x)=cx with c>0.Fig. 1The flow geometry. The flow geometry. In the following physical scenario, assuming the previously stated assumptions, solutal equations, the continuity, thermal, momentum are as follows [[55], [56], [57], [58]]:(1)ux+vy=0,(2)ρhnf(uux+vuy)=μhnfuyy−μhnfK*u+πJ0M08exp(−πpy),(3)(ρcp)hnf(uTx+vTy)=(khnf+16σ*T∞33k*)Tyy+(T−T∞)Q0,(4)(vCy+uCx)=DBCyy−kr(TT∞)m(−C∞+C)exp(−EaKT),(5)vNy+uNx=DnNyy−bWc(Cw−C∞)(NyCy+NCyy), Boundary conditions [[55], [56], [57], [58]]:(6)v=Dm1−CwCy,u=uw,T=Tw,N=Nw,C=Cw,aty=0,(7)C→C∞,u→0,N→N∞,T→T∞,aty→∞. Similarity transformation(8)u=f′(η)cx,v=−cνff(η),(9)η=cνfy,(10)Θ(η)=N−N∞Nw−N∞,θ(η)=T−T∞Tw−T∞,(11)ϕ(η)=C−C∞Cw−C∞, The subsequent are reduced PDEs into ODEs:(12)W1F″′+W2(FF″−F′2)−W1KF′+Qexp(−Aη)=0,(13)1Pr(W3+Rd)θ″+W4(−θ′F)+QTθ=0,(14)ϕ″−Sc(2Fϕ′)−ScRc(1+δθ)mexp(−E1+δθ)ϕ=0.(15)Θ″−Pe((Ω+Θ)ϕ″+Θ′ϕ′)−Se(FΘ′)=0, BCs(16)F′=1,θ=1,Θ=1,f=SbScϕ′,ϕ=1,(17)F′→0,ϕ→0,θ→0,Θ→0. Activation energy parameter E=EakτT, Prandtl number Pr=vρcpkf, Hartman number Q=πJ0M08ρC2, porousity parameter K=νfkpC, Stefan blowing factorSb=Cw−C∞1−Cw,Sc=νfDB Schmidt number, δ=(Tw−T∞)T∞ temperature difference, chemical reaction parameterRc=c2Kr2νf,QT=Q0c(ρCp)fheat source factor. The skin friction, mass transmission, local rate of heat is reflected in dimensionless form as:(18)h=xqmDB(−C∞+Cw),Cf=τwρfuw2,(19)Nu=xqnkf(−T∞+Tw),Nm=xqnDm(Nw−N∞),(20)τw=μhnf∂u∂y|y=0,qn=−(khnf+16σ*T∞33k*)∂T∂y|y=0,(21)qv=−Dm∂N∂y|y=0,qm=−DB∂C∂y|y=0.(22)Cf(Rex)−0.5=F″(0),(23)Sh(Rex)−0.5=−ϕ′(0),(24)Nu(Rex)−0.5=−θ′(0),(25)Nm(Re)−12=−Θ′(0).(26)W1=μhnfμf,W2=ρhnfρf,W4=khnfkf,W3=σhnfσf,W5=(ρCp)hnf(ρCp)f. Xue thermal conductivity model [9](27)knfkf=ln⁡k3+kf2kf(k3k3−kf)1−φ1+2φ1ln⁡k3+kf2kf(kfk3−kf)1−φ1+2φ1.(28)khnfknf=ln⁡k2+knf2knf(k2k2−knf)1−φ2+2φ2ln⁡k2+knf2knf(knfk2−knf)1−φ2+2φ2, Methodology There are several significant benefits of using the HAM (Homotopy Analysis Method) to resolve nonlinear differential equations. Unlike many conventional perturbation methods, it offers a very flexible analytical approximation strategy that does not rely on tiny or large parameter assumptions. An auxiliary convergence-control parameter in HAM enables the solution series to be adjusted in order to guarantee and speed up convergence. It provides considerable flexibility in selecting initial guesses and auxiliary linear operators, and it can effectively handle high nonlinearities. It is especially helpful for complicated fluid flow and heat transfer problems because it produces explicit series answers that can be methodically refined to attain high accuracy. Artificial neural network- Bayesian regularization strategy The anticipated solution and the reference solution from the proposed model generated by the artificial neural network (ANN) are graphically compared in this section. The ANN results are further assessed through the use of fitness curves, regression plots, and a number of performance metrics, such as the scatter plot of targets (ST), mean squared error (MSE), and error histogram (EH). The ANN model trained with the Bayesian regularized back-propagated algorithm also has its convergence statistics summarized in Table 1.Table 1Results of ANN-BRS for fluid model scenarios I–VI.CaseMSEPerformanceGradientEpochMuTimeTrainingTesting13.6880 × E−132.0758 × E−113.3954 × E−129.64 × E−81515001324.3796 × E−129.8856 × E−111.8914 × E−114.32 × E−91695001131.8290 × E−119.3522 × E−112.9458 × E−119.82 × E−81215001049.8298 × E−131.3503 × E−102.0881 × E−119.94 × E−81775001758.2017 × E−125.7353 × E−111.5688 × E−119.77 × E−81305001264.6911 × E−121.0722 × E−91.6454 × E−109.57 × E−820850022Fig. 2Artificial neural network diagram. Table 1Results of ANN-BRS for fluid model scenarios I–VI.CaseMSEPerformanceGradientEpochMuTimeTrainingTesting13.6880 × E−132.0758 × E−113.3954 × E−129.64 × E−81515001324.3796 × E−129.8856 × E−111.8914 × E−114.32 × E−91695001131.8290 × E−119.3522 × E−112.9458 × E−119.82 × E−81215001049.8298 × E−131.3503 × E−102.0881 × E−119.94 × E−81775001758.2017 × E−125.7353 × E−111.5688 × E−119.77 × E−81305001264.6911 × E−121.0722 × E−91.6454 × E−109.57 × E−820850022 Results of ANN-BRS for fluid model scenarios I–VI. Artificial neural network diagram. The results of the Bayesian regularization backpropagation strategy (ANN-BRS) that was designed and applied to the hybrid nanofluid fluid model are presented in detail in Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, Fig. 8(a&b), which correspond to six different scenarios with the labels I through VI. These situations test the suggested ANN-BRS framework's resilience and adaptability by varying important physical parameters. The convergence behavior of validation, training, and testing datasets over epochs for each scenario is shown in Figs. 3a through 8a. In particular, the model was trained to forecast how the flow gradient F′(η) would behave when the Sb and K changed, the thermal field θ(η) when the thermal parameters Rd changed, and the solutal field ϕ(η) when the E changed, microbes profile Θ(η) when the Peclet number E and the Se change. The convergence curves make it abundantly evident that the ANN-BRS model retains stability and smooth learning dynamics during training. For scenarios I through VI, optimal performance was attained at particular epochs: 151, 169, 121, 177, 130, and 208, respectively. With respective Mean Squared Error (MSE) values of 3.3954×10−12−12,1.8914×10−11,2.9458×10−11,2.0881×10−11,1.5688×10−11, and 1.6454×10−10, the values were exceptionally low. The corresponding training times for these outcomes were 13, 11, 10, 17, 12, and 22 s. These low MSE values demonstrate how well the suggested ANN-BRS solver learns and generalizes the intricate non-linear behaviours of the hybrid nanofluid model. Additionally, the model's stable convergence patterns and declining MSE trends show that it not only fits the training data well but also generalizes well to new data. Figs. 3b to 8b, which show how training parameters like the gradient and the adaptive learning rate factor (Mu) change over time, provide additional evidence of the model's dependability. In every instance, overfitting was prevented by ending the training either according to the gradient's magnitude or after a predetermined number of validation checks. Gradients were found to range from 4.32×10−9 to 9.94×10−8, with Mu values staying at 500 in every instance. These numbers show that a local minimum of the error surface was continuously approached during the training process, creating an extremely stable and ideal result. Figs. 3c to 8c display fitness curves that show the degree of contract among the exact or reference solutions obtained from the model and the predicted ANN solutions. Across all parameter scenarios, the neural network has effectively captured the fundamental dynamics of the system, as demonstrated by the overlay of the predicted and reference curves. There are insignificant approximation errors because the smallest deviations fall within acceptable numerical boundaries. Figs. 3d to 8d offer error histograms that illustrate the difference between the target and predicted values in order to further evaluate the error distribution. The fact that most of the error values fall into a small group around zero indicates that the ANN-BRS model consistently and accurately generates predictions. For each of the six scenarios, the average error bin values were 9.82×10−7,−4.2×10−7,8.64×10−7,1.05×10−6,−9.9×10−7, and 7.64×10−6, in that order. The model's predictive power is finally set by the regression plots in Figs. 3e to 8e, which show the connection among the actual target values and network's outputs. A solid line indicates the best-fit regression line, and a dashed line represents the ideal scenario, in which outputs equal targets, in each plot. In every instance, the correlation coefficient (R-value) was found to be very near to 1, indicating that the actual and predicted results had a nearly perfect linear relationship. This high correlation illustrates that the ANN-BRS model is able of accurately signifying the physics behind the hybrid nanofluid flow and thermal-concentration distributions under different physical conditions. The ANN model, which is based on Bayesian Regularization, is a very efficient, accurate, and the graphical and numerical findings provide a solid analytical framework for resolving the nonlinear differential equations associated with the HNF model. The showed ANN-BRS technique is solid and dependable for both theoretical and practical applications in complex fluid dynamics, as evidenced by its near-perfect regression analysis, accurate fitness performance, smooth convergence behavior, and consistently low MSE values. Figs. 8(a&b) and (c&d) demonstrates the velocity profiles on fluid and dust phase rises in tandem with an upsurge inSb. This method is called "Stephan blowing." It involves addition extra energy to a liquid to make it move faster. The fluid gains accelerate and energy when a result of the additional energy source. The temperature field decline while velocity field rise as augments the values of Stephan blowing factor. Fig. 9a&b show the result of E on ϕ(η) field 2D and 3D The activation energy rises when the Arrhenius function is changed. When liquid particles are activated, they produce greater heat energy. The speed of the diffusion process rises. The width of the BL also increases as the E values do. A lower reaction rate physically slows down a chemical process. which is the result of a lower temperature and a larger activation energy. As the fluid concentration distribution shrinks, a chemical reaction parameter rises. The concentration field upsurges as we rise the value of E. The results of Pe and Se on Θ(η) and Nm are demonstrated in Fig. 10(a&b), Fig. 10(c&d) and 9(c&d). An upsurge in the Peclet number (Pe) decline in microorganism dispersion inside the hybrid nanofluid. The cell swimming velocity Wc​ shows a direct relationship with Pe, while both parameters vary inversely with the microbial diffusivity Dn​. Since Pe characterizes the relative dominance of advection over diffusion, higher Pe values intensify advective transport, thereby accelerating microbial flux. As a result, motile microorganisms migrate more rapidly toward the surface, causing a thinning of the microbial boundary layer. Consequently, elevated Pe enhances the local solutal of bacteria near the surface. The importance of φ on Nu, Rd on Nu and Sb on Nu field were demonstration in Figs. 11(a&b). The local rate of heat transportation rise as enhances the values of, Rd and Sb. The Impact of E on Sh are show in Fig. 12(c&d). The local rate of mass transport increase as upsurge the values of AE factor. The rate of mass transmission augment as rise the heat generations factors. The heat transport rates of the previous work with those of the present work are show in Table 2. The two results are clearly reliable with each other, which build up the legitimacy of the current work.Fig. 3Graphical illustrations of ANN-BRS model outcomes for scenario I.Fig. 4Graphical illustrations of ANN-BRS model outcomes for scenario II.Fig. 5Graphical illustrations of ANN-BRS model outcomes for scenario III.Fig. 5aGraphical illustrations of ANN-BRS model outcomes for scenario IV.Fig. 6Graphical illustrations of ANN-BRS model outcomes for scenario V.Fig. 7Graphical illustrations of ANN-BRS model outcomes for scenario VI.Fig. 8(a&b)Significance of Sb on F′.Fig. 8(c&d)Effect of Sb on θ(η).Fig. 9(a&b)Impact of E on ϕ(η).Fig. 9(c&d)Impact of Pe on Θ(η).Fig. 10(a&b)Significance of Se on Θ(η).Fig. 10(c&d)Impact of Pe on Nm.Fig. 11(a&b)Effect of φ on Nu.Fig. 12(c&d)Impact of Rd on Nu.Fig. 13(a&b)Impact of Sb on Nu.Fig. 13(c&d)Impact of E on Sh.Table 2Comparison table.NuPrCurrent results Algehyne et al. [55]0.721.08881.08841.01.33381.33333.02.50992.509610.4.78734.7968Table 3The basic aspects of HNF (hybrid nanofluid) [59,60].CharacteristicsHNμhnfJ1=μhnfμf=1(−φ1+1)2.5(−φ2+1)2.5ρhnfJ2=ρhnfρf=[ρs2ρfφ2+(−φ2+1){(−φ1+1)+φ1ρs1ρf}](ρcp)hnfJ6=(ρcp)hnf(ρcp)f=[(−φ2+1){(1−φ1)+(ρcp)s1(ρcp)fφ1}+φ2(ρcp)s2(ρcp)f]Table 4Properties of based fluid and considered NPs (nanoparticles) [59].+φ2=φProperties constituentscpkρMgo (s1)955453560Water41790.363997.1Ag (s2)23542910,500 Graphical illustrations of ANN-BRS model outcomes for scenario I. Graphical illustrations of ANN-BRS model outcomes for scenario II. Graphical illustrations of ANN-BRS model outcomes for scenario III. Graphical illustrations of ANN-BRS model outcomes for scenario IV. Graphical illustrations of ANN-BRS model outcomes for scenario V. Graphical illustrations of ANN-BRS model outcomes for scenario VI. Significance of Sb on F′. Effect of Sb on θ(η). Impact of E on ϕ(η). Impact of Pe on Θ(η). Significance of Se on Θ(η). Impact of Pe on Nm. Effect of φ on Nu. Impact of Rd on Nu. Impact of Sb on Nu. Impact of E on Sh. Table 2Comparison table.NuPrCurrent results Algehyne et al. [55]0.721.08881.08841.01.33381.33333.02.50992.509610.4.78734.7968 Comparison table. Table 3The basic aspects of HNF (hybrid nanofluid) [59,60].CharacteristicsHNμhnfJ1=μhnfμf=1(−φ1+1)2.5(−φ2+1)2.5ρhnfJ2=ρhnfρf=[ρs2ρfφ2+(−φ2+1){(−φ1+1)+φ1ρs1ρf}](ρcp)hnfJ6=(ρcp)hnf(ρcp)f=[(−φ2+1){(1−φ1)+(ρcp)s1(ρcp)fφ1}+φ2(ρcp)s2(ρcp)f] The basic aspects of HNF (hybrid nanofluid) [59,60]. Table 4Properties of based fluid and considered NPs (nanoparticles) [59].+φ2=φProperties constituentscpkρMgo (s1)955453560Water41790.363997.1Ag (s2)23542910,500 Properties of based fluid and considered NPs (nanoparticles) [59]. Conclusions Consider the outcome of heat generation and Stefan blowing on gyrotactic microorganisms in bio-convection flow across a Riga plate with activation energy. An indication of the outcomes is shown below:❖When we upsurge the values of Stefan blowing factor basis the velocity fields to increases while thermal profile decline.❖Stefan blowing effects in gyrotactic microorganisms in bio-convection flow across a RP (Riga plate) with AE can be reliably and efficiently analyzed and predicted using artificial intelligence techniques, especially Bayesian Regularization Neural Networks (BRNNs). The model enhances the accuracy and depth of the flow analysis by effectively incorporating complex physical phenomena, such as activation energy.❖The solutal field is augmented by an upsurge in the activation energy parameter.❖The Nusselt number at the surface tends to rise as thermal radiation and Stephen blowing parameter parameters increase.❖The value of Sherwood number rises by increasing activation energy parameter using Xue model.❖Increased bioconvection Schmidt and Peclet numbers lead to decreased spread of microorganisms in hybrid nanofluid in both models. Thermal performance of micro devices including biosensors, micromixers, microbial fuel cells, and microfluidic platforms is enhanced by integrating gyrotactic microbes into nanoparticle-based systems. This method increases the efficiency of wastewater treatment, increases the generation of biofuel in photo bioreactors, and makes it possible to create nanostructured materials with consistent thermal characteristics. ❖When we upsurge the values of Stefan blowing factor basis the velocity fields to increases while thermal profile decline. When we upsurge the values of Stefan blowing factor basis the velocity fields to increases while thermal profile decline. ❖Stefan blowing effects in gyrotactic microorganisms in bio-convection flow across a RP (Riga plate) with AE can be reliably and efficiently analyzed and predicted using artificial intelligence techniques, especially Bayesian Regularization Neural Networks (BRNNs). The model enhances the accuracy and depth of the flow analysis by effectively incorporating complex physical phenomena, such as activation energy. Stefan blowing effects in gyrotactic microorganisms in bio-convection flow across a RP (Riga plate) with AE can be reliably and efficiently analyzed and predicted using artificial intelligence techniques, especially Bayesian Regularization Neural Networks (BRNNs). The model enhances the accuracy and depth of the flow analysis by effectively incorporating complex physical phenomena, such as activation energy. ❖The solutal field is augmented by an upsurge in the activation energy parameter. The solutal field is augmented by an upsurge in the activation energy parameter. ❖The Nusselt number at the surface tends to rise as thermal radiation and Stephen blowing parameter parameters increase. The Nusselt number at the surface tends to rise as thermal radiation and Stephen blowing parameter parameters increase. ❖The value of Sherwood number rises by increasing activation energy parameter using Xue model. The value of Sherwood number rises by increasing activation energy parameter using Xue model. ❖Increased bioconvection Schmidt and Peclet numbers lead to decreased spread of microorganisms in hybrid nanofluid in both models. Thermal performance of micro devices including biosensors, micromixers, microbial fuel cells, and microfluidic platforms is enhanced by integrating gyrotactic microbes into nanoparticle-based systems. This method increases the efficiency of wastewater treatment, increases the generation of biofuel in photo bioreactors, and makes it possible to create nanostructured materials with consistent thermal characteristics. Increased bioconvection Schmidt and Peclet numbers lead to decreased spread of microorganisms in hybrid nanofluid in both models. Thermal performance of micro devices including biosensors, micromixers, microbial fuel cells, and microfluidic platforms is enhanced by integrating gyrotactic microbes into nanoparticle-based systems. This method increases the efficiency of wastewater treatment, increases the generation of biofuel in photo bioreactors, and makes it possible to create nanostructured materials with consistent thermal characteristics. Future work: The Levenberg–Marquardt algorithm-based ML model for flow of HNF with oxytactic microbes and melting heat transmission has potential applications in complex thermal control and engineering systems. It can be applied to advance heat transmission efficiency in energy systems where HNF boost efficiency, solar thermal collectors, microelectronics cooling, and biomedical transport processes. In order to improve the design of chemical reactors, drug delivery systems, and environmental flow management, the model can also be used to anticipate complicated nonlinear behaviors in microbe-induced convection and phase-change environments. In the future, these data-driven methods can improve the accuracy of simulating coupled nanofluid–bioconvection events under various physical conditions while drastically lowering computational costs. CRediT authorship contribution statement Waleed Mohammed Abdelfattah: Supervision, Methodology, Conceptualization. Munawar Abbas: Methodology, Formal analysis. Mustafa Bayram: Methodology, Formal analysis. Farkhod Rakhmonov: Methodology, Data curation. Humaira Kanwal: Writing – review & editing. Ali Hasan Ali: Writing – review & editing. Conflict of interest The authors have no relevant financial or non-financial interests to disclose.